presents the percentages or means, standard deviations, and sample sizes for the control, predictor, mediator, and outcome variables by poverty group status. All variables were normally distributed with adequate skew and kurtosis. Zero-order correlations among the variables are displayed in for the poor and near-poor groups. As expected, significant associations were found across the constructs of interest for both poverty status groups, suggesting that family environment, sustained attention, and children’s school readiness outcomes are interrelated. There were some differences by poverty status group. Maternal lack of hostility was positively associated with focused attention only in the near-poor group, whereas maternal stimulation was positively associated with focused attention only in the poor group. Lack of impulsivity was negatively associated with externalizing in the poor group only.
Descriptive Statistics for Control, Predictor, Mediator, and Outcome Variables
Correlations Among Family Environment, Sustained Attention, and School Readiness
Data analysis involved a two-stage process. In the first stage, we identified unique associations between the family environment and the two facets of sustained attention, which was the first aim of our study. In the second stage, we regressed school readiness outcomes on the family environment and both facets of sustained attention. This analysis addressed the second and third aims of our study, namely to identify associations between sustained attention and school readiness and to test whether sustained attention mediated associations between the family environment and school readiness. Given that the last aim of this study was to explore differential associations among these factors between poor and near-poor participants, all analyses were conducted separately by poverty status group.
Regression analyses were consistent with Baron and Kenny’s (1986)
four criteria for mediation. Thus, for mediation to be supported, the following conditions had to be satisfied: (a) family environment should predict school readiness, (b) family environment should predict attention, (c) attention should predict school readiness, and (d) the relation between the family environment and school readiness should be reduced or eliminated when both family environment and attention were entered together into the model. Separate hierarchical regression models were run for the two measures of school readiness, receptive vocabulary and externalizing behaviors. In Step 1 of the model, the school readiness outcome was regressed on the six measures of family environment and all controls. In Step 2, children’s focused attention and lack of impulsivity were added as independent predictors. If the coefficient for a family environment measure decreased from Step 1 to Step 2, a Sobel test (Sobel, 1982
) was conducted to determine whether attention mediated the association between that measure of family environment and that outcome. Both focus and lack of impulsivity were considered as possible mediators. Formal mediation tests were conducted in STATA using sgmediation
, as suggested by Dearing and Hamilton (2006)
. This program uses bootstrap analyses to estimate the indirect effect of the predictor variable on the dependent variable through the mediator variable. Bootstrap analysis involves drawing a large number of samples (with replacement) from a data set, computing the indirect effect for each sample, and then generating an average indirect effect across all samples.
As previously noted, all regression models included controls for child gender, low birth weight, infant temperament, maternal age at first birth, maternal cognitive ability, maternal race/ethnicity, maternal education, marital status, household child:adult ratio, and city. The number of cases with valid values varies across the two school readiness outcomes and is therefore indicated in tables.
The Association Between Family Environment and Sustained Attention
Ordinary least squares regressions were conducted to explore the longitudinal associations between the family environment at 3 years of age and children’s focused attention and lack of impulsivity at 5 years of age. Of particular interest was whether the six facets of the family environment (maternal warmth, maternal lack of hostility, maternal stimulation, physical environment, maternal depression, and maternal parenting stress) would predict attention independent of one another. Thus, measures of attention were regressed simultaneously on all measures of the family environment plus the controls listed above. Focused attention and lack of impulsivity were evaluated in separate models that included the other attention measure as a covariate.
For the near-poor group, lack of maternal hostility (β =.11, p < .05) significantly predicted children’s focused attention, independent of the other five family environment variables (see ). In addition, the link between the physical environment (β = .09, p < .10) and focused attention was marginally significant. The other measures of the family environment did not predict focused attention for this group. Results from analyses predicting lack of impulsivity suggest marginal associations with maternal depression and parenting stress among the near-poor. The other measures of the family environment did not predict lack of impulsivity for this group. For the poor group, none of the family environment variables predicted focused attention or lack of impulsivity. Therefore the second criterion for mediation (an association between family environment and attention) was not satisfied, and formal testing for mediation was not necessary in this group.
Family Environment as a Predictor of Children’s Sustained Attention
Associations Among Family Environment, Sustained Attention, and School Readiness for the Near-Poor
Hierarchical regressions were conducted to determine whether children’s family environment at 3 years of age and attention at 5 years of age were associated with school readiness at 5 years of age. For the near-poor group, as shown in Step 1 of , maternal lack of hostility (β = .14, p < .01) was significantly associated with receptive vocabulary, such that children whose mothers displayed less hostility scored higher on receptive vocabulary. In addition, the physical environment (β = .12, p < .01) reached significance, indicating that children with higher scoring physical environments had larger receptive vocabularies.
Results of Hierarchical Models Predicting School Readiness Outcomes for the Near-Poor
In Step 2, children’s focused attention (β = .32, p < .001), but not lack of impulsivity (β = .06, ns), was strongly associated with receptive vocabulary. Specifically, focused attention accounted for 8% of the unique variance in children’s receptive vocabulary, F(1, 417) = 52.04, p < .001 (not shown). Given that maternal lack of hostility had significantly predicted children’s focused attention (see ), a Sobel test was conducted to formally test whether focused attention mediated the association between lack of hostility and children’s receptive vocabulary. The Sobel test was significant (z = 2.27, p < .05; results not shown) and indicated that focused attention mediated 25.25% of the total effect of maternal lack of hostility on children’s receptive vocabulary. Although the physical environment predicted receptive vocabulary, because it had been only marginally associated with focused attention, a formal test for mediation was not performed. Last, maternal stimulation and maternal parenting stress were marginally significant predictors of receptive vocabulary.
Similar analyses were conducted using externalizing behaviors as the outcome variable. As shown in Step 1 of , maternal parenting stress (β = .20, p < .001) was the only aspect of the family environment that predicted externalizing behaviors. Neither facet of sustained attention emerged as a significant predictor of externalizing behaviors. Therefore, it was not possible for attention to mediate the association between maternal parenting stress and externalizing behaviors in this group.
Associations Among Family Environment, Sustained Attention, and School Readiness for the Poor
The results from a hierarchical regression model predicting receptive vocabulary from family environment and attention within the poor group are displayed in . The results from Step 1 show that maternal stimulation (β =.10, p < .05) predicted receptive vocabulary, such that children with more stimulating mothers had larger receptive vocabularies. The physical environment (β = .09, p < .05) also reached significance, with higher scoring children having larger receptive vocabularies. Finally, children whose mothers scored higher on parenting stress scored lower on receptive vocabulary (β = −.13, p < .001).
Results of Hierarchical Models Predicting School Readiness Outcomes for the Poor
In Step 2, both focused attention (β = .32, p < .001) and lack of impulsivity (β =.14, p < .001) significantly predicted receptive vocabulary. Specifically, attention accounted for 11% of unique variance in receptive vocabulary, F(2, 469) = 53.31, p < .001. Thus, children with better attention demonstrated higher levels of receptive vocabulary in the poor group. However, because none of the family environment factors had predicted either facet of attention, it was not possible that attention mediated associations between the family environment and receptive vocabulary.
Similar analyses were performed to examine the influence of the family environment and attention on poor children’s externalizing behaviors (see ). The results from Step 1 show that maternal warmth (β = −.13, p < .01) predicted externalizing behaviors, such that children with warmer mothers demonstrated fewer externalizing behaviors. In addition, children whose mothers scored higher on parenting stress (β = .17, p < .001) scored higher on externalizing behaviors. Maternal depression was marginally associated with externalizing behaviors.
In Step 2, children’s lack of impulsivity (β = −.09, p < .01), but not focused attention (β = .03, ns), was significantly associated with externalizing behaviors. Children who scored higher on lack of impulsivity had fewer externalizing behaviors. Lack of impulsivity accounted for 1% of the unique variance in children’s externalizing behaviors, F(1, 468) = 3.88, p < .05 (not shown). Given that none of the family environment factors had predicted lack of impulsivity in this group, mediation analyses were not conducted.