مسیر هایی از خشونت زناشویی تا تنظیم احساسات نوزاد: توسعه خروج در مراحل ابتدایی

Abstract Associations between marital conflict and infant emotion regulation exist, but explanatory pathways have not been explored. For older children, parental behavior partially mediates this association through a “spillover” process. We test: associations between mothers’ and fathers’ verbally aggressive marital conflict, infant temperament, and infant withdrawal; mediating effects of negative maternal behavior, and moderating effects of infant temperament, exposure to marital arguments, and contact with father. Eighty mothers, 73 fathers, and their 6-month-old infants participated; parents reported marital aggression prenatally, mothers reported infant exposure to arguments, direct caregiving by father, and infant temperament at 5 months. Negative maternal behavior, infant withdrawal, distress to novelty, activity, and look away were observed at 6 months. Mothers’ and fathers’ aggressive marital conflict predicted infant withdrawal, interactively with exposure to marital arguments and extent of father caregiving, as did infant temperament and negative maternal behavior. Maternal behavior did not mediate between marital conflict and withdrawal.

3. Results Data analyses proceeded in several steps. First, correlations were calculated among all predictors (prenatal maternal depressive symptoms, mothers’ and fathers’ marital aggression, negative maternal behavior, extent of father caregiving, infant exposure to marital arguments, and several temperament variables) and between each predictor and infant withdrawal. Second, withdrawal was regressed on its predictors in a nested set of analyses. In the final analyses, all significant predictors identified in previous analyses were combined in a single, model-testing regression to determine whether mother, father, and infant effects explained non-overlapping variance in infant withdrawal. This approach reduced the number of predictors included in each regression equation and served to maintain an adequate subject to variable ratio in the analyses. According to Harris (1985), an adequate sample size in multiple regression is 50 plus number of predictors, which in these analyses ranged from 10 to 15, indicating that an n of 80 is sufficient. 3.1. Zero-order correlations As presented in Table 2, mother's prenatal depressive symptoms correlated positively with mothers’ marital aggression, and were included as a covariate in subsequent analyses. Mothers’ and fathers’ prenatal marital aggression correlated significantly and positively with each other and with infant exposure to marital arguments; mothers’ marital aggression correlated positively with infant look away behavior. Observed distress to novelty correlated negatively with infant activity, identifying observed distress as a necessary covariate when testing interactive temperament effects. Additionally, observed infant activity and look away correlated positively, indicating covariation and suggesting that both be included in analyses testing the moderating effects of infant temperament (distress to novelty and emotion regulation behavior). Table 2. Zero-order correlations between variables Variable 1 2 3 4 5 6 7 8 9 10 11 1. Maternal depressive symptoms – .31** .10 .08 .14 .10 .09 .04 .02 −.11 .01 2. Mothers’ marital aggression – .48** .13 −.08 .30** .06 .13 .01 .22* .29** 3. Fathers’ marital aggression – .04 .01 .28* −.03 .01 .08 .01 .08 4. Negative maternal behavior – .03 .13 .11 .46** −.15 −.02 .50** 5. Extent of father caregiving – −.11 −.14 −.04 .18 −.03 .04 6. Exposure to marital arguments – .18 .23* −.16 −.08 .32** 7. Reported distress to novelty – .12 −.04 −.23* .05 8. Observed distress to novelty – −.27* −.18 .71** 9. Observed infant activity – .31** −.32** 10. Observed infant look away – −.05 11. Infant withdrawal – Table options Mothers’, but not fathers’, marital aggression correlated positively and significantly with infant withdrawal, as did infant exposure to marital arguments, negative maternal behavior, and observed infant distress to novelty; observed infant activity correlated negatively with withdrawal. No marital variable correlated significantly with negative maternal behavior, thus eliminating the possibility that maternal behavior mediated the association between marital conflict and infant withdrawal. However, observed infant distress to novelty and exposure to marital arguments correlated positively with each other, identifying elevated infant distress to novelty as a possible mediator between exposure to marital arguments and infant withdrawal. 3.2. Hierarchical multiple regressions In view of the number of hypothesized interactions and the sample size, we tested hypothesized interactive effects of each partner's marital aggression with each potential moderator (i.e., infant distress to novelty (temperament), infant exposure to marital arguments, and, for fathers, extent of caregiving with infant) in separate hierarchical regression analyses. For all significant interactions, variables were centered and the interactions were plotted using procedures recommended by Aiken and West (1991). Maternal prenatal depression was a covariate in all analyses to reduce any bias in maternal reporting and collinearity associated with this pre-existing maternal characteristic. 3.2.1. Moderating effects of distress to novelty In the first set of analyses, infant withdrawal was regressed on maternal depressive symptoms, mothers’ marital aggression, and fathers’ marital aggression, followed by observed infant activity and observed look away as potential moderators of distress to novelty, mother-reported and observed infant distress to novelty, and negative maternal behavior entered simultaneously, and then the four 2-way interactions: mother-reported distress to novelty by observed infant activity was included to test the predictive power of this joint measure of infant reactivity to novelty (as discussed above); mother-reported distress by observed infant look away behavior was included to test the moderating effect of the availability of attention shifting, another regulating behavior, on easily distressed infants’ use of withdrawal to regulate distress to novelty. The interactions of reported distress to novelty with marital aggression and with negative maternal behavior were included to test the moderating (exacerbating) effects of infant temperament on the associations between marital and maternal behavior and infant withdrawal. As shown in Table 3, mothers’ marital aggression and maternal negative behavior significantly predicted infant withdrawal at entry and continued to do so after all other single variables had entered the equation, indicating that they explained non-overlapping variance in infant withdrawal from novelty. The distress to novelty interaction (reported distress × activity) was positively and significantly associated with infant withdrawal; the reported distress to novelty × look away interaction was a trend at entry. As shown in Fig. 1, reported distress to novelty was positively associated with observed infant withdrawal only when observed infant activity was high. When infant activity was low, the association was negative, reflecting the significant and negative correlation of infant activity and infant withdrawal. As shown in Fig. 2, and as hypothesized, reported distress to novelty was negatively associated with withdrawal when infants looked away frequently from the novel toy, whereas it was positively associated when infants looked away infrequently. Table 3. Hierarchical multiple regression testing moderating effects of infant temperament (N = 80) Predictors 1 (β) 2 (β) 3 (β) 3 (B) 4 (β) 4 (B) 1. Maternal depressive symptoms −.10 −.09 −.12 −.14 Mothers’ marital aggression .36** .22* .23* .49 Fathers’ marital aggression −.09 −.02 −.03 −.03 2. Observed infant activity −.15t −.82** −.86** Observed infant look away .03 .70t .65 Reported distress to novelty −.04 .38 .56 Observed distress to novelty .56** .58** .57** Negative maternal behavior .21* .16t .21 3. Reported distress × activity .73* .38* .78* .40* Reported distress × look away −.85t −.53t −.80 −.50 4. Reported distress × negative maternal behavior −.05 −.35 Reported distress × mothers’ marital aggression −.02 −.19 Note: R2 = .10* for step 1; ΔR2 = .51** for step 2; ΔR2 = .03t for step 3; ΔR2 = .00 for step 4. tp < .10; *p < .05; **p < .01; β is standardized beta; B is unstandardized beta and reported only for interactions. Table options The association between mother-reported distress to novelty and observed infant ... Fig. 1. The association between mother-reported distress to novelty and observed infant withdrawal varies as a function of observed infant activity. Figure options The association between mother-reported distress to novelty and observed infant ... Fig. 2. The association between mother-reported distress to novelty and observed infant withdrawal varies as a function of infant look away behavior. Figure options Marital aggression remained significant and negative maternal behavior remained a strong trend after entry of the temperament interactions (step 3), demonstrating that they explain primarily non-overlapping variance in infant withdrawal. Observed distress to novelty was strongly associated with infant withdrawal at each step in the equation, indicating that infants who became more distressed during exposure to the novel toy engaged in more withdrawal. Thus, in subsequent regression analyses, distress × activity and observed distress were included as predictors of withdrawal. Given the difficulty of detecting moderation in non-experimental research (McClelland & Judd, 1993) and the power limitations of the current sample, the trend interaction between reported distress to novelty and look away was also included in subsequent analyses in an attempt to understand the effect of correlated interaction terms on the model. Contrary to hypothesis, reported distress to novelty interacted with neither mothers’ nor fathers’ marital aggression, nor with negative maternal behavior to predict withdrawal. To explore the possibility introduced by the significant distress to novelty by activity interaction that this measure of infant temperament would moderate the main effect of marital aggression or of negative maternal behavior on infant withdrawal, we recalculated the analyses, adding the three-way interactions (distress to novelty × activity × marital aggression or distress to novelty × activity × maternal negative) on the last step of the equations. The two-way interactions between marital or maternal behavior and infant activity needed to test the three-way interactions were also included. No significant three-way effects or trends were observed, all p values >.30; this was the case also for interactions of infant temperament with father's marital aggression. 3.2.2. Moderating effect of infant exposure to marital arguments In the second set of analyses, infant withdrawal was regressed on maternal depressive symptoms, mothers’ marital aggression, and fathers’ marital aggression entered first, followed by observed infant activity and look away behavior, reported distress to novelty, negative maternal behavior, and infant exposure to marital arguments entered simultaneously, then the interaction of infant exposure and mothers’ marital aggression. To test the hypothesized mediating effect of observed distress to novelty (on the association between the marital aggression × exposure interaction and infant withdrawal), observed distress to novelty was entered next,4 and its impact on the interaction noted. The two interactions (reported distress to novelty by activity and by look away) were entered on the final step to determine if temperament and exposure effects explained independent variance in infant withdrawal. The interactive effect of fathers’ marital aggression by infant exposure was tested in a separate analysis in which all the same variables, except the interaction of mothers’ marital aggression by infant exposure, were entered as above. As shown in Table 4, the mothers’ marital aggression effect, significant at entry, was qualified by its significant interaction with exposure to conflict. As shown in Fig. 3, prenatal marital aggression was more strongly (and positively) associated with infant withdrawal when infants were exposed to marital arguments postnatally than when they were not. Table 4. Hierarchical multiple regression testing moderating effects of infant exposure to marital arguments (N = 80) Predictors 1 (β) 2 (β) 3 (β) 3 (B) 4 (β) 5 (β) 5 (B) 1. Maternal depressive symptoms −.10 −.12 −.16 −.10 −.13 Mothers’ aggression .36** .28* −.44 −.02 .02 Fathers’ aggression −.09 −.10 −.07 −.04 −.05 2. Observed infant activity −.20* −.16t −.12 −.80* Observed infant look away −.05 −.13 .00 .74t Reported distress to novelty −.06 −.06 −.06 .42 Negative maternal behavior .43** .44** .22** .17t Exposure to conflict .19t −.58 −.13 −.08 3. Mothers aggression × exposure 1.22* .75* .39 .34 4. Observed distress to novelty .52** .54** 5. Reported distress × activity .73* .38* Reported distress × look away −.93t −.58t Note: R2 = .10* for step 1; ΔR2 = .32** for step 2; ΔR2 = .04* for step 3; ΔR2 = .17** for step 4; ΔR2 = .03* for step 5. tp < .10; *p < .05; **p < .01; β is standardized beta; B is unstandardized beta, reported only for interactions. Table options The association between mothers’ marital aggression and observed infant ... Fig. 3. The association between mothers’ marital aggression and observed infant withdrawal varies as a function of infant exposure to marital arguments. Figure options Consistent with the hypothesized mediating effect,4 the interaction of mothers’ marital aggression and infant exposure to marital arguments was no longer significant after observed infant distress entered the equation (β dropped from 1.22, p < .01–.39, n.s.), whereas observed distress continued to predict infant withdrawal. The interaction of reported distress to novelty by activity was significant on the final step after entry of the marital aggression by exposure interaction, and the interaction of reported distress to novelty by look away remained a trend, indicating that they explained variance in infant withdrawal that did not overlap with variance explained by the marital aggression by infant exposure interaction. When the regression was recalculated substituting fathers’ marital aggression by infant exposure for the maternal aggression interaction, the effect was a trend, β = 1.24, p = .06, and operated as above, with increasing marital aggression associated with increasing infant withdrawal when infant exposure to marital arguments was high. Consistent also with the effect for mothers’ marital aggression, this interactive effect became non-significant after entry of observed distress (β dropped from 1.22, p < .10–.04, n.s.), indicating again that the effect of exposure to fathers’ aggressive marital conflict on infant withdrawal is mediated by the infant's negative arousal. All other effects were as reported above using mothers’ marital aggression. The fathers’ marital aggression by infant exposure interaction was not included in the final model because it was a trend, and because the two interactions explained shared variance in withdrawal. 3.2.3. Moderating effect of father caregiving In the third set of analyses, infant withdrawal was regressed on maternal depressive symptoms, mothers’ marital aggression, and fathers’ marital aggression, followed by observed activity and look away behavior, reported distress to novelty, negative maternal behavior, and father caregiving entered simultaneously, then the father marital aggression × caregiving interaction to test the hypothesized moderating effect, followed by observed infant distress, and then the temperament (reported distress × activity, reported distress × attention) interactions to determine if temperament and father marital effects were independent. As hypothesized, fathers’ marital aggression interacted with paternal caregiving to predict infant withdrawal after all single variables had entered the equation, as shown in Table 5. As illustrated in Fig. 4, when fathers engage in high amounts of infant caregiving, fathers’ marital aggression is positively associated with infant withdrawal, whereas when father caregiving is low, a negative association occurs between fathers’ aggression and infant withdrawal. This interactive effect was reduced to a trend when the temperament interactions entered the equation, and the temperament effects were similarly reduced. Thus, we infer that the interactive effects of infant temperament and caregiving by maritally aggressive fathers are predominantly, though not entirely, independent. Table 5. Hierarchical multiple regression testing moderating effects of father caregiving (N = 80) Predictors 1 (β) 2 (β) 3 (β) 3 (B) 4 (β) 4 (B) 5 (β) 5 (B) 1. Maternal depressive symptoms −.10 −.14 −.11 −.09 −.21 Mothers’ marital aggression .36** .33** .35** .25** .26** Fathers’ marital aggression −.09 −.06 .42* −.29t −.26t 2. Observed infant activity −.25** −.27** −.18* −.74* Observed infant look away −.05 −.08 .02 .62 Reported distress to novelty −.01 .00 −.01 .38 Negative maternal behavior .43** .44** .21** .17* Extent of father caregiving .11 −.74t −.51 −.44 3. Fathers’ aggression × father caregiving .95* .03* .69* .02* .61t .02t 4. Observed distress to novelty .54** .56* 5. Reported distress × observed activity .61t .31t Reported distress × observed look away −.77 −.48 Note: R2 = .10* for step 1; ΔR2 = .30** for step 2; ΔR2 = .04* for step 3; ΔR2 = .21** for step 4; ΔR2 = .02 for step 5. tp < .10; *p < .05; **p < .01: β is standardized beta; B is unstandardized beta, reported only for interactions. Table options The association between fathers’ marital aggression and observed infant ... Fig. 4. The association between fathers’ marital aggression and observed infant withdrawal varies as a function of the extent of father caregiving. Figure options 3.2.4. Test of the complete model In the final regression analysis, infant withdrawal was regressed on maternal depressive symptoms, mothers’ marital aggression, and fathers’ marital aggression entered first, followed by observed infant activity and look away behavior, reported infant distress to novelty, negative maternal behavior, father caregiving, and infant exposure to marital arguments entered simultaneously, the previously identified marital aggression interactions (i.e., mothers’ marital aggression × infant exposure to marital arguments; fathers’ marital aggression × father caregiving), followed by observed infant distress,4 and then the temperament interactions (reported distress × activity; reported distress × look away). As shown in Table 6, maternal behavior was significant at entry, and remained significant after all simple effects and interactions had entered the equation, as shown in step 5, demonstrating its robustness as a predictor of infant withdrawal and its independence from mothers’ and fathers’ marital aggression and infant temperament. The interactions (of mothers’ marital aggression × infant exposure and fathers’ marital aggression × father caregiving) explained significant and non-overlapping variance in infant withdrawal from novelty. When the temperament variables, including observed infant distress, entered the equation, the mother aggression interaction was no longer significant (as shown also in Table 4) and the father aggression interaction was reduced to a trend; the temperament interactions were trends as well. Table 6. Hierarchical regression: effects of fathers’ and mothers’ marital aggression and infant temperament on withdrawal Predictors 1 (β) 2 (β) 3 (β) 3 (B) 4 (β) 4 (B) 5 (β) 5 (B) 1. Maternal depressive symptoms −.10 −.14 −.16 −.11 −.13 Mothers’ marital aggression .36** .29* .40 .00 .02 Fathers’ marital aggression −.09 −.10 −.41* .29t −.27t 2. Observed infant activity −.23* −.20* −.16* −.71* Observed infant look away −.04 −.15 −.01 .67t Reported distress to novelty −.04 −.03 −.03 .43 Negative maternal behavior .42** .44** .23** .18* Extent of father caregiving .12 −.66t −.47 −.40 Exposure to conflict .20t −.58t −.14 .11 3. Fathers’ aggression × father caregiving .88* .03* .66t .02t .57t .02t Mothers’ aggression × exposure to conflict 1.21* .75* .41 .25 .38 .23 4. Observed distress to novelty .50** .52** .44** 5. Reported distress × observed activity .60t .31t Reported distress × observed look away −.85t −.53t Note: R2 = .10t for step 1; ΔR2 = .33** for step 2; ΔR2 = .07** for step 3; ΔR2 = .16** for step 4; ΔR2 = .02t for step 5. tp < .10; *p < .05; **p < .01: β is standardized beta; B is unstandardized beta, reported only for interactions.