SRM Perceiver, Target, and Family Effects in Perceived Problem Behavior
First, we assessed with SRM analysis whether significant perceiver and target variances and a significant family variance could be found in family members’ perceptions of each other’s problem behavior. We tested two models, one for externalizing problem behavior and one for internalizing problem behavior. The fit of the externalizing problem behavior model was acceptable. The chi-square test was significant, χ2 (N = 69, df = 47) = 72.84, p = .01, indicating a lack of fit. However, a RMSEA of .08 and a CFI of .96 indicated that the model could be accepted without modification. For perceptions of internalizing problem behavior, the SRM model also fitted the data adequately. The chi-square test was significant, χ2 (N = 69, df = 47) = 64.62, p = .04, but a RMSEA of .06 and the CFI of .97 indicated an adequate fit.
Both models were analyzed again approximately ten months later to determine if the findings would replicate. In the second wave, the initial fit of the model for perceived externalizing problem behavior was not adequate. The modification indices indicated that allowing the residuals of fathers’ ratings on the problem behavior of the older and younger adolescents to covary would improve the fit of the model. This adjustment did not alter the significance of effects and only slightly changed their variance. The adjusted model resulted in χ2 (N = 69, df = 46) = 64.73, p = .04. The RMSEA was .07 and the CFI .97, which indicated that there was an adequate fit for the model. The fit of the model for perceived internalizing problem behavior did not differ much from the first wave: χ2 (N = 69, df = 47) = 64.46, p = .05, RMSEA of .04 and CFI of .96. Overall, these values suggest that the SRM fits the data on perceptions of internalizing problem behavior reasonably well in the second wave.
The variances of the SRM effects for perceived externalizing and internalizing problem behavior in both the first and second wave are shown in Table
2. The significance of the between-family variance in the various SRM effects was tested with a one-tailed T-test. As shown in the columns labelled “First Wave, Externalizing Problems” and “Second Wave, Externalizing Problems”, we found that all four perceiver and target effects (for fathers, mothers, older adolescents, and younger adolescents) and the family effect explained significant amounts of variance in perceptions of externalizing problem behavior in the first wave as well as in the second wave. This indicates that within-family perceptions of externalizing problem behavior are due to the way fathers, mothers, older adolescents, and younger adolescents perceive externalizing problem behavior in their family members (i.e., perceiver effects for fathers, mothers, older adolescents, younger adolescents); the degree to which fathers, mothers, older adolescents, and younger adolescents elicit perceptions of externalizing problem behavior in their family members (i.e., target effects for fathers, mothers, older adolescents, younger adolescents); and the general level of externalizing problem behavior that family members perceive among each other (i.e., family effect). Our hypothesis that perceiver and target effects and the family effect could be found in family members’ perceptions of each other’s problem behavior is thus supported for perceptions of externalizing problem behavior.
Table 2
Variance estimates for the social relations model analyses of perceived externalizing and internalizing problem behavior: perceiver, target, and family effects
Perceiver | Father | 0.042*** | 0.031*** | 0.039*** | 0.037*** |
Mother | 0.065*** | 0.030*** | 0.056*** | 0.029*** |
Adolescent 1 | 0.085*** | 0.078*** | 0.071*** | 0.038*** |
Adolescent 2 | 0.068*** | 0.083*** | 0.071*** | 0.076*** |
Target | Father | 0.005* | 0.003 | 0.014** | 0.012** |
Mother | 0.010** | 0.010** | 0.007* | 0.004 |
Adolescent 1 | 0.026*** | 0.035*** | 0.022*** | 0.033*** |
Adolescent 2 | 0.014*** | 0.020*** | 0.006* | 0.018*** |
Family | | 0.018** | 0.011* | 0.018** | 0.012** |
As can be seen from column in Table
2 labelled “First Wave, Internalizing Problems”, one can see that we found four significant perceiver effects (for fathers, mothers, older adolescents, and younger adolescents), three significant target effects (for mothers, older adolescents, and younger adolescents), and a significant family effect in perceptions of internalizing problem behavior in the first wave. The non-significant target effect for fathers indicates that across families, there was no variability in how much fathers are perceived to have internalizing problems. As can be seen in the column labelled “Second Wave, Internalizing Problems”, we also found four reliable perceiver effects (for fathers, mothers, older adolescents, and younger adolescents) and a reliable family effect in perceptions of internalizing problem behavior in the second wave. The variances of the target effects of fathers, older adolescents, and younger adolescents were significant, but the target effect of mothers was not significant. Thus, we found that within-family perceptions of internalizing problem behavior could be attributed to (a) the degree to which fathers, mothers, older adolescents, and younger adolescents perceive internalizing problem behavior in their family members (i.e., the perceiver effects), (b) the degree to which older adolescents and younger adolescents elicit perceptions of internalizing problem behavior in their family members (i.e., the target effects), and (c) the general level of internalizing problem behavior that family members perceive among each other (family effect). Within-family perceptions of the degree to which mothers elicit perceptions of internalizing problem behavior (i.e., target effect mothers) in the first wave did not replicate in the second wave. In the second wave, within-family perceptions of the degree to which fathers elicit perceptions of internalizing problem behavior (target effect fathers) were found that had not been significant in the first wave. For perceptions of internalizing problem behavior, our hypothesis is supported with respect to two target effects (for both adolescents), four perceiver effects (for fathers, mothers, older adolescents, and younger adolescents), and the family effect. Our hypothesis was only partially supported with regard to two target effects (for both adolescents) in perceptions of internalizing problem behavior.
Other findings that were not directly related to our hypotheses emerged from the estimation of individual and dyadic reciprocity correlations. We found some significant individual reciprocity correlations. For externalizing problem behavior, a perceiver-target correlation was found for older adolescents (r = .67, t = 6.64, p ≤ .01 first wave; r = .58, t = 5.36, p ≤ .01 second wave), indicating that older adolescents who perceive more externalizing problem behavior in other family members tend to be perceived by their family members as higher in externalizing problem behavior. In the first wave, we also found a perceiver-target correlation for younger adolescents (r = .29, t = 2.42, p ≤ .05) with respect to externalizing problem behavior. This implies that younger adolescents who perceive more externalizing problem behavior in other family members tend to be perceived by their family members as higher in externalizing problem behavior.
In the second wave, the perceiver-target correlations for older adolescents (r = −0.31, t = −2.58, p ≤ .01) and younger adolescents (r = −.29, t = −2.48, p ≤ .01) were significant for internalizing behavior. Older and younger adolescents who perceive more internalizing problem behavior in other family members tend to be perceived by their family members as lower in internalizing problem behavior. We found no significant dyadic reciprocity correlations for perceptions of externalizing and internalizing problem behavior in either the first or the second wave.
The Relative Contributions of Perceiver, Target, and Family Effects in Perceptions of Externalizing and Internalizing Problem Behavior
The percentage of variance explained by the SRM effects was assessed for each of the 12 family members’ perceptions of each other’s problem behavior. For example, the total variance in fathers’ perceptions of mothers’ externalizing problem behavior is the sum of the perceiver variance for fathers (see Table
2, .042), the target variance for mothers (see Table
2, .010), the family variance (see Table
2, .018), and any remaining unexplained variance (.301, not in Table
2). In this case the sum is .371. The relative perceiver variance in fathers’ perceptions of mothers’ externalizing problem behavior is computed by dividing the fathers’ perceiver variance by the total variance (i.e., .042/.371 = 11.32%). The contributions of the different SRM effects to the variance in perceptions of externalizing problem behavior as well as in perceptions of internalizing problem behavior were next averaged for all 12 family members’ perceptions of each other’s problem behavior. In Table
3, the relative amounts of variance accounted for by the SRM perceiver and target effects and the family effect are presented.
Table 3
The relative amounts of variance in perceptions of externalizing and internalizing problem behavior accounted for by the srm perceiver, target, and family effects
SRM perceiver effects | 30.69 | 25.39 | 24.06 | 20.39 |
SRM target effects | 6.41 | 7.85 | 4.46 | 7.50 |
SRM family effects | 8.64 | 4.81 | 7.27 | 5.03 |
Unexplained/SRM relationship effects | 54.25 | 61.95 | 64.23 | 67.09 |
As can be seen in Table
3, SRM perceiver variance explained 24–31% of the variance in perceptions of externalizing problem behavior and 20–25% in perceptions of internalizing problem behavior. SRM target variance explained 4.5–6.5% of the variance in perceptions of externalizing problem behavior and 7.5–8% in perceptions of internalizing problem behavior. Finally, SRM family variance explained 7–9% of the variance in perceptions of externalizing problem behavior, and 5% of the variance in perceptions of internalizing problem behavior. Our hypothesis that the absolute value of the amount of variance explained in family members’ perceptions of problem behavior by target effects is greater for judgments of externalizing problem behavior than for judgments of internalizing problem behavior was not supported. We did not find evidence for greater agreement between family members in their ratings of externalizing problem behavior compared with their ratings of internalizing problems. Our findings suggest a tendency of greater correspondence in perceptions of internalizing problem behavior.
About 54–64% of the variance in externalizing problem behavior and 62–67% of the variance in internalizing problem behavior is unexplained. Considering that the systematic relationship variance became part of the residual variance (see section “Method”), part of this variance can be accounted for by characteristics of specific relationships between family members (i.e., SRM relationship effects). However, because these components also contain non-systematic (error) variance, we can not draw conclusions about their relative importance. For those components only containing systematic variance, the largest absolute value of the amount of variance explained in family members’ perceptions of problem behavior is accounted for by the SRM perceiver effects. For externalizing behavior, the absolute value of the amount of variance explained by the SRM family effect is slightly greater than the average for the SRM target effects, whereas for internalizing problem behavior the absolute value of the amount of variance explained by the SRM target effects is slightly greater than by the SRM family effect.