Testing for Socially Desirable Responding
Examination of the raw scores from the Marlow-Crowne Form C Social Desirability measure produced a categorical inference of socially desirable responding (whenever five or more responses were socially desirable). No point-biserial correlations between the categories of socially desirable/non-socially desirable responding and the construct index scores were statistically significant. Considering that analysis by point-biserial correlation has lower statistical power than correlations with two continuous variables, correlations with the raw score on the Marlow-Crowne were also examined. There were no statistically significant correlations between the raw score on the Marlow-Crowne and the construct index scores used in these analyses. This result indicates that socially desirable responding does not confound the analysis in this experiment.
Confirmatory Factor Analysis The first step of the analysis was to determine whether the items loaded on the appropriate constructs. This test was necessary to support the validity of the measured variable index scores. The confirmatory factor analysis revealed that items had convergent validity for the appropriate constructs. All item loadings were statistically significant, there were no z scores for the loadings below 10.0 (see Table 1). Nineteen out of the 25 item-factor correlation coefficients were .80 or above, the lowest three were .52, .54, and .62, still in the moderate range. Cronbach�s alpha, indicating the reliability of the scales used for index scores, and correlations between the item and the factor can be seen in Table 1. All scales used to create index scores had a Cronbach�s alpha above .80. The goodness of fit indicators of the confirmatory factor analysis suggested that this model could be improved upon (see Table 6).
There were several significant covariances between factors (see Table 2). As would be expected from the hypothesized model structure, the likelihood of helping significantly covaried with three of the predictors, expectancy for target helping if positions in the scenario were reversed (expectancy), oneness, and empathic concern. Involvement significantly covaried with oneness, empathic concern, sadness, aversive arousal, and expectancy. This reveals that involvement is a significant indicator of general cognitive arousal and that the scenario was effective in producing the desired impact in the respondents. Sadness and aversive arousal were correlated, r2 = .75, p < .000001, indicating a moderate relationship between these adverse reactions hypothesized to be egoistic pathways to helping. Sadness and aversive arousal also correlated with empathic concern and oneness, indicating that egoistic and altruistic motivational pathways are not mutually exclusive. The estimated likelihood that the target would help if positions were reversed (expectancy) significantly covaried with oneness and empathic concern. This may indicate that individuals with relatively stronger relational bonds are expected to be more likely to provide assistance in critical situations. There was also a significant covariance between oneness and empathic concern. This finding replicates Cialdini et al.�s (1997) results and has implications for the relationship between altruistic and egoistic motivations.
Table 1. Structural coefficients for the confirmatory factor analysis
| Oneness | 1: �To what extent would you use the term �we� to describe your relationship with your friend?� | 0.71 | 0.06 | .54 | 11.91 | < .000001 |
| alpha = .8314 | 2: Modified IOS Scale | 1.14 | 0.05 | .93 | 23.86 | < .000001 |
| 3: Convergence representation task | 1.21 | 0.05 | .95 | 24.97 | < .000001 | |
| Empathic Concern | 1: �Moved� | 1.17 | 0.05 | .84 | 21.74 | < .000001 |
| alpha = .9395 | 2: �Compassionate� | 1.17 | 0.05 | .89 | 23.90 | < .000001 |
| 3: �Soft-hearted� | 1.40 | 0.05 | .95 | 26.52 | < .000001 | |
| 4: �Tender� | 1.39 | 0.06 | .89 | 23.89 | < .000001 | |
| Sadness | 1: �Feeling low� | 1.70 | 0.07 | .90 | 23.53 | < .000001 |
| alpha = .8705 | 2: �Low-spirited� | 1.80 | 0.07 | .93 | 24.76 | < .000001 |
| 3: �Heavy-hearted� | 1.47 | 0.09 | .71 | 16.75 | < .000001 | |
| Aversive Arousal | 1: �Disturbed� | 1.43 | 0.07 | .80 | 19.98 | < .000001 |
| alpha = .9025 | 2: �Troubled� | 1.82 | 0.07 | .92 | 24.62 | < .000001 |
| 3: �Uneasy� | 1.73 | 0.07 | .89 | 23.15 | < .000001 | |
| Indebtedness | 1: �If you attempted to save your friend, would you expect them to do the same for you someday?� | 1.28 | 0.10 | .62 | 13.33 | < .000001 |
| alpha = .8036 | 2: �If you helped your friend in this situation, would this obligate them to help you in the future?� | 1.67 | 0.10 | .75 | 16.29 | < .000001 |
| 3: �To what degree would your friend be indebted to you if you tried to rescue him/her?� | 1.99 | 0.10 | .92 | 20.13 | < .000001 | |
| Expectation | 1: �If your positions were reversed, how likely do you think your friend would be to save you?� | 1.31 | 0.06 | .90 | 23.83 | < .000001 |
| alpha = .9171 | 2: �Imagine that you were the one in trouble, to what extent would your friend risk his/her life to try to save you?� | 1.36 | 0.05 | .94 | 25.76 | < .000001 |
| 3: �To what degree would your friend be indebted to you if you tried to rescue him/her?� | 1.25 | 0.06 | .84 | 21.33 | < .000001 | |
| Likelihood of Helping | 1: �How likely is it that you would go in and try to save your friend?� | 1.38 | 0.05 | .95 | 26.48 | < .000001 |
| alpha = .9279 | 2: �To what extent would you risk your own life to try to save your friend?� | 1.14 | 0.05 | .85 | 21.87 | < .000001 |
| 3: �If you were actually in this situation, what is the probability that you would run into the building? (from 0 to 100%)� | 1.41 | 0.06 | .91 | 24.60 | < .000001 | |
| Involvement | 1: �To what degree did you get involved in the story?� | 1.33 | 0.07 | .84 | 18.64 | < .000001 |
| alpha = .7390 | 2: �I did not become very involved in the story.� | 1.39 | 0.09 | .69 | 14.89 | < .000001 |
| 3: �I took an objective perspective when reading the story and did not get caught up in it.� | 1.05 | 0.10 | .52 | 10.53 | < .000001 |
There was some complexity apparent from the results of the confirmatory factor analysis. Theoretically relevant cross-loadings from a second analysis, a model constructed based on the modification indices of the first CFA, are presented in Table 3. The degree to which a respondent would use the term �we� to describe her/his relationship with the target (One1) had a correlation of .16 with empathic concern. Although this indicator of oneness shared 3% of the variance with empathic concern, it shared 29% of the variance with oneness. Surprisingly, the convergence representation task (One3) had a small but statistically significant inverse relationship with empathic concern, r2 = .0025, p < .05. The empathic concern item for �moved� (Emp1) shared 1% of the variance with oneness and 71% of the variance with empathic concern. None of these crossloadings appear to be a severe threat to the divergent validity of the measured constructs of oneness and empathy.
Table 2. Significant correlations between factors in the confirmatory factor analysis
| Oneness and Empathic Concern | 0.30 | 0.05 | 6.56 | < .000001 |
| Sadness and Oneness | 0.11 | 0.05 | 2.08 | < .019 |
| Sadness and Empathic Concern | 0.28 | 0.05 | 6.02 | < .000001 |
| Aversive Arousal and Oneness | 0.15 | 0.05 | 2.99 | < .0014 |
| Aversive Arousal and Empathic ConcernXXXXX | 0.38 | 0.04 | 8.46 | < .000001 |
| Aversive Arousal and Sadness | 0.75 | 0.03 | 29.36 | < .000001 |
| Expectation and Oneness | 0.40 | 0.04 | 9.30 | < .000001 |
| Expectation and Empathic Concern | 0.38 | 0.04 | 8.74 | < .000001 |
| Helping and Oneness | 0.25 | 0.05 | 5.35 | < .000001 |
| Helping and Empathic Concern | 0.34 | 0.04 | 7.49 | < .000001 |
| Helping and Expectation | 0.81 | 0.02 | 39.83 | < .000001 |
| Involvement and Oneness | 0.24 | 0.05 | 4.47 | < .000004 |
| Involvement and Empathic Concern | 0.65 | 0.04 | 18.17 | < .000001 |
| Involvement and Sadness | 0.29 | 0.05 | 5.47 | < .000001 |
| Involvement and Aversive Arousal | 0.37 | 0.05 | 7.46 | < .000001 |
| Involvement and Expectation | 0.41 | 0.05 | 8.63 | < .000001 |
| XXXXXX | XXXXXX | XXXXXX | XXXXXX |
There did appear to be a high degree of complexity for the sadness indicator item for �heavy-hearted� (Sad3). This item targeting sadness shared 50% of the variance with sadness and 24% of the variance with aversive arousal. This finding should be noted for future research. It is possible that this high crossloading is an artifact of instrument design, because the sadness and aversive arousal items alternated in a six-item subsection of the questionnaire. The goodness of fit indicators (see Table 6) revealed that including factorial complexity improved the fit of the model.
Table 3. Theoretically relevant crossloadings in a second confirmatory factor analysis
| One1 on Empathic Concern | 0.08 | 2.65 | .16 | < .0041 | |
| One3 on Empathic Concern | 0.04 | -1.68 | -.05 | < .047 | |
| Emp1 on Oneness | 0.04 | 3.72 | .11 | < .0001 | |
| Sad3 on Aversive Arousal | 0.12 | 8.49 | .49 | < .000001 |
Path Analyses
Model justification. A preliminary causal model tested whether kinship would have a direct effect on helping in the absence of any mediators. This saturated model indicated a significant direct influence of kinship on helping, z = 5.92, p < .000001. This finding encourages testing the mediated effects of kinship hypothesized in this study. The first mediated model (see Figure 4) included oneness, empathic concern, sadness, aversive arousal, expectancy for helping by the target if positions in the scenario were reversed, and obligation after being helped, as proximate mediators of kinship on helping. The model also included the direct influence of kinship on helping. Results from this analysis were used to test hypotheses 1 through 5. Interaction terms for kinship by oneness, empathy, expectancy, and indebtedness were added to the model to test hypotheses 6 and 7.
Hypothesis 1: The impact of kinship on the predictors of helping. The prediction that oneness, empathic concern, sadness, and aversive arousal would be higher for sibling targets than for friend targets was not supported (see Table 4). In fact, oneness was significantly lower for siblings than for friends, z = -4.06, p < .000026. No other differences based on the manipulation of kinship were found in the mean values of these mediators.
Hypothesis 2: The role of psychological experiences in helping. Sadness, aversive arousal, and empathic concern significantly covaried with oneness (See Table 2). The prediction that empathic concern would not make a statistically significant unique contribution to the variance in the likelihood of helping, once the effects of kinship, reciprocal altruism, and oneness were accounted for, was not supported. Results indicated that empathic concern accounted for 10% of the variance in the likelihood of helping, z = 7.03, p < 000001 (see Table 4). As predicted, aversive arousal and sadness did not make a statistically significant unique contribution to the variance in the likelihood of helping, once other factors were accounted for. As predicted, oneness made a statistically significant unique contribution to the variance in the likelihood of helping, z = 8.09, p < 000001, accounting for 12% of the variance in this model (see Table 4).
Hypothesis 3: The unique contribution of kinship to the likelihood of helping. The prediction that kinship would make a significant unique contribution to the amount of variance explained in the likelihood to help was supported. As hypothesized, this effect was positive and statistically significant, z = 5.65, p < .000001 (see Table 4). This result indicated that kinship has an effect on the likelihood of helping, even when established influences such as empathy and oneness have been accounted for. The direct effect of kinship accounted for 5% of the variance in the likelihood of helping. In a post-hoc analysis, kinship was inversely correlated with respondents� ratings of liking for, r(444) = -.36, p < .01 (two-tailed), and similarity to, r(444) = -.21, p < .01 (two-tailed), the target character (ratings were lower for siblings than for friends). These factors are not likely to be mediators of kin selecting influences.
Figure 4. Baseline mediational model with standardized coefficients
Chi-square(6) = 4.16, p > .65
*3 indicates p < .001
*5 indicates p < .00001
*6 indicates p < .000001