Procedure

Participants received verbal instructions on the general purpose and nature of the experiment. After informed consent was obtained, participants completed one form of the inventory. Participants were asked to think of a person they knew (a sibling who is close in age or a close friend they have known for several years) and give the first name, gender, and age of this person on the questionnaire form. Cialdini et al. (1997) mentioned the usefulness of items for priming an appropriate focus on the target character. In order to facilitate this focus, participants responded to 18 items asking about their similarity with and degree of liking for the target. These items were presented in randomized order.

Participants then read a 500-word scenario (with a Flesch-Kincaid grade level of 6.4) in which they were described to be with the target character at a hotel in the Rocky Mountains. Blank spaces appeared in the story when the target character was mentioned, participants were asked to fill in the name of the target when they encounter a blank space. At the end of each scenario, there was a possibility that the target was in a life-threatening situation. One previous study found kin selecting effects when the labels �A good friend� and �Your sibling� were used for target characters, regardless of whether the participants were known to have siblings or good friends (Kruger, In Press a). Another previous study found kin selecting effects when the labels �Your brother Mike� and �Your friend Mike� were used for target characters, regardless of whether these individuals were known to exist in reality (Kruger, In Press b). The use of target characters actually known to the participants and the repetitive task of filling in the target�s name enhances the mundane realism of the study, and is expected to promote kin-selecting effects.

Respondents completed items assessing the likelihood of making a risky life-saving attempt, perspective taking, self-other overlap, empathic concern, aversive arousal, sadness, reciprocity, and involvement. Filler sections were used to reduce demand characteristics and test for bias in response. These included the Marlowe-Crowne, Form C Social Desirability measure (Silverstein, 1983) an attachment style measure and a measure of dispositional empathy. Each participant responded to two scenarios with identical sets of items, one for sibling targets and one for friend targets, in a randomized order. Only the data from the first scenario was used in the analysis.

Analytical Strategy

Confirmatory factor analysis. Analyses of the hypothetical framework utilized path modeling based on measured variable index scores. The first step in the data analysis was to verify the items� construct validities. Items would need to load strongly on the hypothesized factor in order to be considered valid indicators suitable for creating index scores. A Confirmatory factor analysis (CFA) tested the extent to which the items loaded on the appropriate constructs. Because one creates index scores by averaging across several items, a small factor loading would indicate that a particular raw item score would introduce considerable error into the measured variable index. A second CFA was conducted based on the results of the first CFA. The second CFA allowed for factorial complexity in some of the items, as suggested by the modification indices from the first CFA. This procedure permits the examination of items that may load on more than one factor, to determine their utility for the present and future research. Items with comparable cross-loadings may not be suitable for an analysis that seeks to partial causal influences.

Path model of measured variables. This between-subjects analysis used data from the set of items for the first scenario. Multiple items were used to create an index score for each measured variable. SPSS 9.0 compiled the dataset and computed measured variable scale scores. LISREL 8.3 created the hypothesized models and performed the data analysis using the covariance matrix among the measured variables. A path model of measured variable index scores was constructed from the dataset (see Figure 3). This model examined the influence of the participants' mental experiences (empathic concern, oneness, sadness, aversive arousal) and beliefs about reciprocity (target's likelihood of helping if positions were reversed, target's expected indebtedness to the donor after being helped) on the likelihood of helping. The model also examined the influence of kinship with the target on the psychological predictors of helping behavior (empathic concern, etc.), and the unique effect of kinship on the likelihood of helping. Examination of this model provided tests of Hypotheses 1 through 5.

Procedure for testing hypotheses 1-5. Hypotheses 1 through 5 were tested by the examination of coefficients for the path model displayed in Figure 3. Assigning a value of 0 to data from friend scenarios and 1 to data from sibling scenarios facilitated the testing of Hypothesis 1. Positive path coefficients for gamma 2,1, gamma 3,1, gamma 3,1, gamma 4,1, gamma 5,1 and gamma 6,1, would indicate that the means for these variable indices (and thus the psychological constructs) were higher for siblings than for friends in the path model. The paths from sadness (beta 1,4), aversive arousal (beta 1,3), and empathic concern (beta 1,5) to the likelihood of helping provided a test of Hypothesis 2. Statistically significant path coefficients would indicate that these factors play a role in the likelihood of helping.

The path coefficient (gamma 7,7) between kinship and the likelihood of helping provided a test of Hypothesis 3. A positive and statistically significant value for this path coefficient would indicate that genetic relationship made a significant unique contribution to the amount of variance explained in the likelihood of helping. Examination of the path coefficients for the influence of obligation after being helped (beta 1,1) and the expectancy for helping by the target in a role reversal (beta 1,2) on the likelihood of helping tested Hypothesis 4 in the same manner. These coefficients indicated whether or not cognitions related to reciprocal altruism made an impact on the likelihood of helping. A statistically significant negative path coefficient for gamma 7,1 and a significantly positive path coefficient for gamma 7,2 would support Hypothesis 5. This would reveal that respondents perceived less indebtedness for siblings to reciprocate than for friends.

Figure 3.Path model of measured variable index scores

Procedure for testing the hypotheses 6-7. Interaction terms were created to test Hypotheses 6 and 7 by centering the continuous variables and dummy coding for kinship. Multiplying the values of 0 for friend data and 1 for sibling data by scores on the psychological predictors of helping (oneness and empathic concern) and the constructs relevant to reciprocal altruism (indebtedness of the target after being helped and expectations for target helping if positions in the scenario were reversed), created four separate interaction terms, which were added to the model as direct predictors of the likelihood of helping. Positive values for these coefficients would indicate that the effect (of oneness, etc.) was higher for siblings than for friends. Rather than testing whether the level of oneness experienced was greater for siblings than for friends, this method tested whether the degree of oneness would have a greater impact on the likelihood of helping a sibling than a friend.

Hypothesis 6 predicted that reciprocity would have a greater influence in helping for friends than for siblings, because friends do not benefit from kin selecting tendencies. The kinship x expectancy (expectation that the target would help if positions in the scenario were reversed) and kinship x indebtedness interaction term were expected to have statistically significant negative path coefficients. Hypothesis 7 predicted that oneness and empathic concern would be stronger predictors of helping for siblings than for friends. The coefficients for the kinship x oneness and kinship x empathic concern interactions terms were predicted to be positive and statistically significant.

Developing a more parsimonious model. If a path coefficient is not statistically significant, then eliminating the path will not significantly worsen the fit of the model. If a path can be removed without reducing the fit of the model, then the effect measured by that path is not significant (see Kline, 1999). The elimination of nonsignificant paths creates a more parsimonious model explaining the influences on altruistic helping behaviors. This model provides a new baseline for testing competing models of altruistic helping. The kinship interaction terms were not considered when evaluating the goodness of fit in this model because they are not actually measured variables but linear combinations of variables already in the model.

Testing Alternative Models. Other research groups have developed theoretical frameworks that predict alternative conceptualizations of the constructs in this experiment. Path analyses tested models based on Batson et al. (1997) and Cialdini et al.�s (1997) conceptualization of psychological predecessors to helping behaviors. The fit of these models were compared to the fit of the parsimonious model constructed based on the integration of evolutionary and social psychological theories. The parsimonious and earlier models developed for this study hypothesized that psychological predictors of helping behaviors, such as empathy, serve as mediators of the effects of kinship on helping. These models proposed a simple one-step mediational pathway. Alternative models with mediational complexity were also tested. The fit of these more complex models were tested against the fit of the parsimonious model.

Testing the effects of respondent gender and target gender. An additional mediational model was created which includes the effects of respondent gender and target gender (see Figure 7). Due to the nature of the scenario, where respondents could make a risky attempt to rescue someone who may be in a hazardous situation, it was predicted that male respondents would be more likely to help in general than female respondents. At the same time, female respondents are expected to experience higher levels of empathy, sadness, and aversive arousal, based on the results of previous research (Brody, 1993). This may result in a mediated pathway that increases the likelihood of helping by raising levels of empathy (see Trobst, Collins, and Embree, 1994), however this effect is not expected to be as strong as the unique effect of respondent gender. It was also predicted that female targets would be more likely to be rescued than male targets, due to the �heroic rescuing� qualities of the scenario. Tests of these hypotheses involved observations of the appropriate path coefficients in the model including gender effects, in the manner previously described for testing the effects of kinship.