خودکارآمدی فرزند پروری و مشکلات رفتاری در کودکان در معرض خطر بالا برای مشکلات رفتاری اولیه: نقش واسطه ای افسردگی مادر

Abstract Parenting self-efficacy (PSE) has been positively linked to children's adjustment and negatively associated with maternal depression. However, most PSE research has been cross-sectional, limited to predominantly white, middle-class samples, and has not examined potential mechanisms underlying associations of PSE with children's behavior. The present study investigates: (1) how PSE changes over time, (2) the relationship between age 2 PSE and children's behavior problems 2 years later, and (3) the potential mediating role of maternal depression in relation to the association between PSE and child problem behavior. Participants are 652 ethnically and geographically diverse mothers and their children, at high risk for conduct problems. PSE increased between ages 2 and 4 and higher initial levels predicted lower caregiver-reported age 4 conduct problems after controlling for problem behavior at age 2. The relationship between PSE and later conduct problems was mediated, however, by maternal depression. These findings suggest maternal depression as a potential disruptor of caregiver confidence in early childhood, which has implications for the design and focus of parenting interventions.

. Introduction The relationship between parenting and children's problem behavior has been well-documented (Dishion & Patterson, 2006; Gardner, Ward, Burton, & Wilson, 2003). Given the principle of hierarchical integration, it is generally true that successful parenting in early childhood sets that stage for the same in middle childhood and adolescence, to the beneficence of the developing young person. Research on parenting in early childhood confirms that parents play a more prominent role in children's socialization process relative to older ages (Campbell, Shaw, & Gilliom, 2000; Deater-Deckard, 2000, Fagot, 1997 and Gardner, 1987; Pettit & Bates, 1990; Shaw & Bell, 1993). An important component of human behavior, and change and adaptation of behavior, is the sense of efficacy to complete tasks or execute complex skills (Bandura, 1977 and Bandura, 2006). This principle has been applied to parenting, with the development of measures and studies of parenting self-efficacy (PSE). Perceived competence in the parenting role has been positively linked with observed parenting competence (e.g., warmth, sensitivity, & engagement). In fact, PSE has been found to mediate associations between parenting and children's developmental outcomes (Coleman & Karraker, 2003). In light of this promising research on PSE, it is somewhat surprising to find a dearth of literature on the developmental course or stability of PSE or longitudinal data linking its development with children's socioemotional outcomes. In addition, most research on PSE has been conducted with predominantly European American, middle-class families (see Raver & Leadbeater, 1999; Zayas, Jankowski, & Mckee, 2005 for notable exceptions). Even fewer studies have examined potential mediating factors that might account for associations between PSE and child problem behavior. While PSE has been found to mediate associations between parenting and child outcomes, it is possible that underlying intrapersonal characteristics such as parental depression might contribute to PSE and its association with child outcomes. The current study sought to advance our knowledge on PSE by examining its developmental course from the toddler to preschool period, by exploring its longitudinal relation with emerging children's conduct problems, and testing whether associations between PSE and children's conduct problems were potentially mediated by maternal depressive symptoms. The study was carried out with a large sample of 652 children at high risk for developing early conduct problems based on the presence of socioeconomic, family, and child risk factors. 1.1. Self-efficacy and parenting PSE has been defined as the degree to which parents expect to competently and effectively perform their roles as parents (Teti & Gelfand, 1991), and it is rooted in general self-efficacy theory. Guided by social learning theory, general self-efficacy refers to the belief in one's ability to perform behaviors successfully (Bandura, 1977). Overall, self-efficacy includes the motivation, cognitive resources, and courses of action necessary to implement control over a specific task or event (Ozer & Bandura, 1990). Self-efficacy reflects perceived self-competence as opposed to expectations of task success or failure (Bandura, 1977 and Bandura, 2006). In other words, an individual with high self-efficacy may anticipate task failure in a situation that would realistically require advanced expertise in a specific domain. In general, high levels of self-efficacy has been found to predict competence in the face of environmental demands, conceptualize difficult situations as challenges, have less negative emotional arousal in the face of stress, and exhibit perseverance when challenged (Jerusalem & Mittag, 1995). In contrast, low self-efficacy is associated with self-doubt, high levels of anxiety when faced with adversity, assuming more responsibility for task failure than success, interpreting challenges as threats, and avoiding difficult tasks. Given that parenting in early childhood is characteristically complex and filled with change and unpredictability (Shaw & Bell, 1993), it would seem that the parent's sense of efficacy would be germane to understanding which parents will rise to the occasion, or alternative, which will become more discouraged and perhaps deteriorate over time, leading to increasing problem behavior in young children. 1.2. The developmental course of parenting self-efficacy We know relatively little about its developmental course during early childhood, a period of time that has been found to be critical for parenting and children's subsequent adjustment (Shaw, Bell, & Gilliom, 2000). During toddlerhood, children rapidly acquire a repertoire of cognitive, social, and motor skills, challenging parents to tailor their parenting techniques to children's changing needs. In turn, children during this key developmental period are especially dependent on the influence of their caretakers in terms of their social and emotional adjustment. Because of the unique circumstances presented during this period of development not only for children, but also for the growth of parents, it is critical to understand how PSE unfolds specifically during early childhood. One study followed changes in PSE among low-income minority mothers from the third trimester of pregnancy to 3 months post-partum (Zayas et al., 2005). The authors found that PSE significantly increased during the transition to motherhood. In one of the few other longitudinal studies using an ethnically diverse, middle-class sample, Gross, Conrad, Fogg, and Wothke (1994) examined changes in PSE among two cohorts of children from ages 1 to 2 and 2 to 3, respectively, with measurements of PSE three times each year. PSE was found to increase between ages 1 and 2 in Cohort 1, but remain stable from ages 2 to 3 in Cohort 2. In light of these studies, there is some evidence that PSE initially increases in the first couple of years of children's lives and then shows moderate stability; however, longitudinal modeling of growth parameters across at least three measurement waves is needed to elucidate the developmental trajectory of PSE. 1.3. Parenting self-efficacy and children's socioemotional adjustment As early-onset behavior problems have been linked to the development of more severe conduct problems in middle childhood and adolescence, such as delinquency and substance use (Campbell et al., 2000; Shaw & Gross, 2008), it is critical to identify factors in early childhood that contribute to or protect children from the maintenance of problem behavior. Based on PSE's theoretical and empirical links to several dimensions of parenting behavior (Bor & Sanders, 2004; Coleman et al., 2002), it is logical to consider its associations with problem behavior during early childhood. In fact, PSE has been linked to children's development in terms of behavioral adjustment (Bor & Sanders, 2004). For example, as early as 5 months, PSE was found to be positively related to concurrent ratings of infant soothability (Leerkes & Crockenberg, 2002). At age 2, Raver and Leadbeater (1999) found that PSE was inversely related to children's concurrent difficult temperament among a sample of urban impoverished families. Furthermore, children's observed compliance, negativity, and avoidance of mother at age 2 was found to be associated with concurrent ratings of PSE among predominantly middle-class, mother–toddler dyads (Coleman & Karraker, 2003). Among a demographically similar sample of mothers with school-aged children, higher PSE was concurrently associated with less emotionally reactive and more sociable behavior (Coleman & Karraker, 2000). In terms of problem behavior, lower levels of PSE among mothers of preschool-aged children at high risk for developing conduct problems were found to be associated with higher levels of concurrent children's disruptive behaviors (Bor & Sanders, 2004). Furthermore, mothers of clinically referred, 2- to 8-year-old children with conduct problems reported lower levels of PSE than a comparative community sample (Sanders & Woolley, 2004). Despite the existence of several studies examining relations between PSE and different dimensions of child functioning, most work in this area has been cross-sectional. Furthermore, with notable exceptions (Raver & Leadbeater, 1999; Zayas et al., 2005), most work on PSE has been conducted using predominantly white, middle-class samples (e.g., Coleman et al., 2002; Leerkes & Crockenberg, 2002; Teti & Gelfand, 1991). Thus, more research that incorporates culturally and socioeconomically diverse samples is needed to extend previous findings to under-represented groups. 1.4. Maternal depression as a mediator In addition to needing more studies that trace the development course of PSE over time and link PSE to specific child outcomes using samples of children at risk for high rates of clinically meaningful psychopathology, few researchers have attempted to examine potential mediators of associations between PSE and child adjustment. The development of general self-efficacy is thought to result from four domains of experiences, including individual histories of task successes versus failures, vicariously learning through the successes and failures of others, verbal persuasion from others, and aversive physiological arousal, such as stress responses (Bandura, 1989). While PSE has been theoretically linked to a number of early life experiences (e.g., history of childhood maltreatment, maternal stress), maternal depression represents one important intrapersonal filter through which the detrimental effects of maternal PSE on children's functioning may be transmitted. Specifically, hallmarks of depression such as feelings of helplessness and worthlessness may underlie and drive feelings of low maternal PSE and lead to the development of children's problem behavior. Thus, associations between low PSE and children's conduct problems may be accounted for by the negative cognitions and affect associated with depression. As both clinical depression and subclinical elevated depressive symptomatology have been associated with children's maladjustment (Cummings, Keller, & Davies, 2005; Farmer, McGuffin, & Williams, 2002), the terms maternal depressive symptoms and depression are used throughout this paper to describe elevated symptoms that were measured on a continuous scale. To build a logical case for considering maternal depression as a mediator between PSE and child conduct problems, it is important to establish covariation between PSE and child problem behavior. In fact, maternal depression has been found to be associated with both low PSE (Bor & Sanders, 2004; Haslam, Pakenham, & Smith, 2006; Teti & Gelfand, 1991; Zayas et al., 2005) and children's problem behavior, including conduct problems (Owens & Shaw, 2003; Shaw, Keenan, & Vondra, 1994). For example, Bor and Sanders (2004) found that maternal depressive symptoms were negatively related to concurrent PSE. However, as most research on the relationship between maternal depression and PSE has been cross-sectional, it is difficult to determine the precise direction of effects. Overall, studies examining associations between maternal depression and different facets of child adjustment have been more prevalent, including longitudinal studies examining associations with child conduct problems carried out in early childhood. While not uniformly consistent and sometimes suffering from informant response bias (Fergusson, Lynskey, & Horwood, 1993) by relying on one informant for reports of both maternal depression and child problem behavior, the pattern of overall results suggest a longitudinal association between maternal depression in early childhood and both preschool and later school-age conduct problems (Shaw et al., 2000). In terms of establishing the link between PSE and maternal depression, research on general self-efficacy and negative life events considered to be dependent on maternal behavior (e.g., divorce, change in residence, and loss of job) has suggested that lower self-efficacy is more prevalent among mothers with histories of depression (Maciejewski, Prigerson, & Mazure, 2000). These findings indicate that maternal depression could play an important role in affecting the course of PSE, and ultimately affecting the impact of low PSE on children's problem behavior. Given the established link between maternal depression and child problem behavior, particularly conduct problems (Owens & Shaw, 2003; Shaw et al., 1994), we sought to test the possibility that maternal depression may serve as an underlying mechanism linking PSE to children's conduct problems. 1.5. Present study The present study had three primary goals. First, we sought to characterize the trajectory of PSE over three time points during early childhood among a large sample of families facing high levels of socioeconomic, family, and child risk. Second, associations between PSE and later child conduct problems were explored, controlling for age 2 child problem behavior and using multiple reporters of child problem behavior 2 years after initial assessment of PSE. A third goal was to explore the role of maternal depressive symptoms at age 3 as a potential mediator of the covariation between PSE and subsequent child conduct problems while controlling for the autoregressive effect of maternal depression at age 2. Effects of maternal race and child's gender on PSE, maternal depression, and child problem behavior were also explored. In comparison to the vast majority of prior research on PSE, the current study, the Early Steps Multisite Project, included a longitudinal design, the use of a large, ethnically diverse sample of children at high risk for clinically meaningful problem behavior, and the exploration of a potential mechanism underlying the relationship between PSE and child conduct problems: maternal depression. Based on the limited longitudinal research on PSE (Gross et al., 1994 and Zayas et al., 2005) and mother's increasing experience in the parenting role, we hypothesized to see increases in PSE over time. Second, based on prior research (Bor & Sanders, 2004; Sanders & Woolley, 2004), we expected PSE to be negatively associated with children's later conduct problems even after accounting for initial conduct problems using both maternal and alternate caregiver (AC) reports (tested as separate structural models). Finally, based on prior cross-sectional studies linking PSE to maternal depression (Haslam et al., 2006 and Zayas et al., 2005) and maternal depression to later child conduct problems (Owens & Shaw, 2003), maternal depressive symptoms at age 3 were expected to mediate associations between PSE at age 2 and both maternal and AC reports of child conduct problems at age 4.

. Results 3.1. Descriptive statistics and intercorrelations Means and standard deviations for study variables are presented in Table 1. It should be noted that T-scores (M = 50, SD = 10) are presented for all measures of children's conduct problems. On average, children were nearly one standard deviation above the normative mean for maternal reports of age 2 externalizing behaviors as measured by the Child Behavior Checklist (M = 59.23, SD = 7.95), age 2 maternal-reported Eyberg intensity of behavior (M = 58.94, SD = 7.95), and maternal-reported Eyberg problem behavior scale at both ages 2 and 4 (M = 59.07, SD = 8.40 and M = 59.72, SD = 10.94, respectively). In terms of maternal depression, scores of 16 and above on the CES-D have been found to reflect clinically meaningful depressive symptoms ( Eaton & Kesslerm, 1981; Myers & Weissman, 1980). Specifically, this cut-off score showed a modest relationship with clinical depression in community samples ( Wetzler & van Praag, 1989). On average at both ages 2 and 3, mothers in the sample reported elevated levels of depressive symptomatology (M = 16.57, SD = 10.39 and M = 15.59, SD = 11.02, respectively), with over 40% scoring at or above the clinical cut point of 16 at each time point. Table 1. Descriptive statistics for measured variables Mean Standard deviation Maternal reports PSE (age 2) 30.31 4.66 PSE (age 3) 30.93 4.84 PSE (age 4) 31.09 4.71 CBCL externalizing behaviors (age 2) 59.23 7.95 Eyberg behavior intensity (age 2) 58.94 7.95 Eyberg behavior problem (age 2) 59.07 8.40 CBCL externalizing behaviors (age 4) 53.71 10.37 Eyberg behavior intensity (age 4) 57.78 9.63 Eyberg behavior problem (age 4) 59.71 10.71 CES-D depression score (age 2) 16.57 10.39 CES-D depression score (age 3) 15.59 11.02 Alternate caregiver reports CBCL externalizing behaviors (age 2) 53.70 9.53 Eyberg behavior intensity (age 2) 55.76 8.20 Eyberg behavior problem (age 2) 50.30 9.54 CBCL externalizing behaviors (age 4) 49.73 9.77 Eyberg behavior intensity (age 4) 54.41 8.17 Eyberg behavior problem (age 4) 51.51 10.17 Table options Intercorrelations of study variables are presented in Table 2. With the exception of the nonsignificant correlations between age 2 PSE and AC-reported Eyberg problem score at age 4 and age 2 maternal depression and AC-reported externalizing problems at ages 2 and 4, and nonsignificant trends between age 2 PSE and AC-reported externalizing problems (r = −0.10, p = .10), and age 2 maternal depression and age 4 AC-reported Eyberg intensity score at age 4 (r = 0.10, p < .10), all variables were significantly correlated in the expected directions. Specifically, PSE scores were positively associated across time, PSE scores were negatively correlated with children's problem behaviors at ages 2 and 4 and maternal depression at ages 2 and 3, and age 3 maternal depression was positively associated with all factors of children's problem behaviors at ages 2 and 4. Table 2. Correlation matrix of measured variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 1. PSE (age 2) 2. PSE (age 3) .60* 3. PSE (age 4) .56* .63* 4. Externalizing (age 2)-M −.22* −.28* −.22* 5. Externalizing (age 2)-AC −.21* −.18* −.19* .30* 6. Intensity (age 2)-M −.13* −.13* −.11* .48* .22* 7. Intensity (age 2)-AC −.17* −.18* −.24* .31* .67* .30* 8. Problem (age 2)-M −.14* −.13* −.13* .38* .25* .59* .21* 9. Problem (age 2)-AC −.23* −.22* −.23* .23* .59* .20* .64* .29* 10. Externalizing (age 4)-M −.18* −.28* −.34* .49* .28* .36* .25* .23* .16* 11. Externalizing (age 4)-AC −.10† −.24* −.26* .30* .36* .17* .34* .14* .24* .39* 12. Intensity (age 4)-M −.18* −.26* −.33* .40* .24* .43* .26* .25* .18* .78* .29* 13. Intensity (age 4)-AC −.16* −.24* −.33* .22* .31* .18* .38* .14* .29* .39* .72* .35* 14. Problem (age 4)-M −.17* −.25* −.30* .38* .25* .29* .23* .35* .25* .68* .32* .75* .35* 15. Problem (age 4)-AC −.07 −.25* −.23* .14* .30* .16* .29* .17* .35* .25* .58* .22* .67* .25* 16. Depression (age 2) −.16* −.14* −.14* .16* .05 .13* .12* .10* .12* .21* .07 .19* .10† .17* .15* 17. Depression (age 3) −.21* −.34* −.29* .24* .13* .17* .17* .12* .18* .32* .23* .28* .23* .26* .28* .44* Note. PSE denotes parenting self-efficacy; M denotes maternal report and AC denotes alternate caregiver report. * p < .05. † p < .10. Table options 3.2. Data analysis approach Latent growth curve modeling (LGCM) was used to model initial levels of, and changes in, PSE over time. LGCM captures individual differences in development longitudinally by using structural equation modeling to estimate the mean intercept (starting point) and slope (rate of change) of individuals based on their observed scores on multiple indicators of a specific construct of interest. Given the scores on the observed variables, maximum likelihood estimates are used to find the most likely values of the unobserved latent growth parameters. The mean latent intercept and slope are then used to describe the shape of the average growth curve (McArdle & Bell, 2000). LGCMs can be used to make predictions about the means and covariances of the data, yielding fit indices including χ2 goodness of fit significance test, the comparative fit index (CFI; Bentler, 1990), Tucker–Lewis Index (TLI; Tucker & Lewis, 1973), root mean square error of approximation (RMSEA; Brown & Cudeck, 1993), and the standardized root mean square residual (SRMR). CFI and TLI values greater than 0.95 and RMSEA and SRMR values less than 0.05 indicate good model fit. A nonsignificant χ2-value indicates that the proposed covariance matrix does not differ from the observed covariance matrix, suggesting adequate model fit. Because the χ2-test is sensitive to sample size ( Widaman & Thompson, 2003), the ratio of χ2/df provides a better fit index with larger sample sizes ( Bollen, 1989), with values less than 2.5 indicating adequate model fit. For all analyses, missing data were handled using full information maximum likelihood estimation in Mplus Version 4.0 ( Muthén & Muthén, 1998). Tests of mediation were conducted using the bootstrap sampling method to test the significance of the indirect path (Shrout & Bolger, 2002). Data are randomly drawn with replacement from the observed dataset to create a new dataset of the same size. This procedure is repeated a specified number of times (1000 for the current study), and the indirect effect is estimated from each of the datasets, yielding a confidence interval. An advantage of the bootstrap method is that it does not assume a normally distributed parameter estimate, which is important given that distributions of indirect effect estimates repeatedly have been found to be positively skewed (Mackinnon, Lockwood, Hoffman, West, & Sheets, 2002; Shrout & Bolger, 2002). As such, when assuming normality, the resulting confidence intervals produce asymmetric error rates, thereby decreasing the power necessary to detect the indirect effect (Mackinnon et al., 2002). Because bootstrap methodology does not assume a normal distribution, it is a more powerful test than traditional multi-step approaches, and is recommended for small to moderate sample sizes (Shrout & Bolger, 2002). 3.3. Unconditional latent growth curve model of PSE 3.3.1. Goal 1: Change in PSE over children's ages 2–4 First, in order to model change in PSE over time, an unconditional (i.e., no covariates were included in the model) LGCM was fit using Mplus Version 4.0 (Muthén & Muthén, 1998). All fit indices indicated that the model was a good fit to the data, χ2(2) = 4.93, p > .05; CFI = 0.99, TLI = 0.99, RMSEA = 0.04, SRMR = 0.03. Parameter estimates of the unconditional LGCM suggested that average initial levels of PSE were significantly different from zero (b = 30.40, SE = 0.17, p < .01); however, this is not particularly meaningful given there was no zero point on the PSE scale that was administered. More importantly, there was significant individual variability around the intercept (b = 13.08, SE = 0.95, p < .01), meaning that individuals significantly deviated from average levels of PSE at age 2. In terms of growth, PSE significantly increased between ages 2 and 4 (b = 0.40, SE = 0.08, p < .01). There was not significant individual variation around the average slope. As a result, predictors and outcomes were only extended to the intercept parameter for all conditional LGCMs. 3.4. Conditional latent growth curve models of PSE 3.4.1. Goals 2 and 3: Covariations between PSE, children's problem behaviors, and maternal depression Separate conditional latent growth curve models were fit for maternal versus AC reports of children's problem behavior. Parameter estimates of both models are presented in Fig. 1 (using maternal reports of problem behavior) and Fig. 2 (using AC reports of problem behavior). In both models, error terms were allowed to correlate between each manifest indicator of children's behavior problems and its counterpart across ages 2 and 4 to account for shared method variance. For example, the age 2 externalizing error term was correlated with the age 4 externalizing error term. Furthermore, error variances between the Eyberg intensity score and the Eyberg problem score were allowed to correlate within time point at ages 2 and 4 to account for shared method variance. For parsimony, error variances were not represented in the models. Latent growth curve model of parenting self-efficacy (PSE) from ages 2 to 4, ... Fig. 1. Latent growth curve model of parenting self-efficacy (PSE) from ages 2 to 4, maternal depression at age 3, and maternal-reported children's problem behaviors at age 4 (showing unstandardized path coefficients; mediational model in bold for emphasis). Note that site and treatment group status are included as covariates, but are not represented in the model for simplicity. Figure options Latent growth curve model of parenting self-efficacy (PSE) from ages 2 to 4, ... Fig. 2. Latent growth curve model of parenting self-efficacy (PSE) from ages 2 to 4, maternal depression at age 3, and alternate caregiver-reported children's problem behaviors at age 4 (showing unstandardized path coefficients; mediational model in bold for emphasis). Note that site and treatment group status are included as covariates, but are not represented in the model for simplicity. Figure options In terms of the LGCM using maternal reports of children's behavior problems, all fit indices suggested that the model was a good fit, χ2(74) = 166.99; p = .00; χ2/df = 2.23, CFI = 0.96; TLI = 0.95, RMSEA = 0.04; SRMR = 0.05. As shown in Fig. 2, lower levels of PSE at age 2 (i.e., intercept) predicted higher levels of maternal-reported children's problem behavior at age 4 (b = −0.38, SE = 0.11, p < .01) after controlling for the significant effects of age 2 levels of problem behavior on both initial levels of PSE (b = −0.21, SE = 0.04, p < .01) and age 4 problem behavior (b = 0.65, SE = 0.08, p < .01), as well as maternal educational attainment on PSE (b = −0.45, SE = 0.14, p < .01) and significant treatment effects on both maternal depression at age 3 (b = −1.91, SE = 0.76, p < .05) and children's behavior problems (b = −1.70, SE = 0.56, p < .01). Site differences were also covaried in the model; however, effects were nonsignificant. Main effects of maternal race and children's gender were also explored. We compared European American mothers versus mothers of racial minority groups (i.e., African American, Biracial, and “other”) due to the small sample sizes of minority groups other than African American. Mothers of racial minority groups were found to have higher initial levels of PSE as compared to European American mothers (b = −0.86, SE = 0.35, p < .05). There were no significant effects of race on maternal depression at age 3 or on age 4 behavior problems. In terms of child's gender, there were nonsignificant trends for mothers with girls to have higher levels of depression at age 3, controlling for age 2 depression (b = 0.19, SE = 0.11, p < .10) and for boys to have higher rates of age 4 problem behavior after controlling for initial levels of conduct problems (b = −0.16, SE = 0.08, p < .10). Significant covariations between initial levels of PSE and maternal depression at age 3 and between age 3 depression and children's behavior problems at age 4 were in the expected direction (b = −.76, SE = 0.13, p < .01 and b = 0.10, SE = 0.02, p < .01, respectively). In terms of the mediating effect of maternal depression at age 3 while controlling for age 2 levels of depression, the confidence interval of the indirect effect based on 1000 bootstrap samples did not include zero (95% CI: −0.13 to −0.03), suggesting significant mediation of maternal depression on the covariation between PSE at age 2 and maternal-reported children's problem behavior at age 4. Next, because the bootstrapping technique is recommended when fitting models with moderate sample sizes and the Sobel test requires more power to detect effects ( Mackinnon et al., 2002), mediation also was tested using the more rigorous, large sample, Sobel method ( Sobel, 1982). Results using the Sobel test were also significant (z = −2.13, p < .05), further supporting the mediating role of maternal depression. Because of the potential reporter bias in the LGCM using only maternal reports, the identical model was fit using AC reports of children's problem behavior at ages 2 and 4. All fit indices suggested that the model was a very good fit to the observed data, χ2(71) = 102.24; p = .01; χ2/df = 1.44; CFI = 0.97; TLI = 0.96; RMSEA = 0.04; SRMR = 0.04. Results showed similar patterns of effects with a significant negative relationship between PSE at age 2 and children's problem behaviors at age 4 (b = −0.32, SE = 0.05, p < .01). Again, significant relationships were found between initial levels of PSE and maternal depression at age 3 while controlling for the autoregressive effect of age 2 depression, and between age 3 maternal depression and children's behavior problems at age 4, both in the expected direction (b = −0.72, SE = 0.25, p < .01 and b = 0.10, SE = 0.04, p < .01, respectively). As compared to the model with only maternal reports, the model with AC-reported children's problem behavior showed nonsignificant pathways between maternal level of education and initial PSE and between age 2 problem behavior and age 3 maternal depression. There were also nonsignificant intervention effects on maternal depression at age 3 and age 4 problem behaviors. There were no significant effects of maternal race on PSE, maternal depression, or children's problem behavior in the AC-report model. In terms of gender, boys were found to have higher levels of age 4 problem behavior than girls, while controlling for initial levels (b = −0.28, SE = 0.11, p < .05). With respect to mediation, the confidence interval of the indirect effect of age 3 depression while controlling for age 2 levels of depression based on 1000 bootstrap samples did include zero (95% CI: −0.17 to −0.01); however, the upper bound suggests a nonsignificant trend, providing some corroborating evidence of the mediating role of maternal depression. Results using the Sobel test of the indirect effect also indicated a trend towards mediation for the AC-report model (z = −1.95, p < .10).