Measuring Effortful Control in Research on Aggression Prevention

Background

Skillful social interaction requires conscious management of emotional responses, which occurs through a process called effortful control.  Often conceptualized as a cognitive skill, effortful control is negatively associated with externalizing problems.  Accordingly, many school-based prevention programs attempt to improve effortful control in children.

Researchers have developed neuropsychological tests and paper-and-pencil instruments to measure effortful control or aspects of it.  However, effortful control has been defined inconsistently, and the psychometric properties of effortful control measures are still emerging.  In this exploratory study we sought to develop a teacher-rated effortful control measure in order to evaluate a social skills training program for elementary school students.

Methods

Fourteen rural North Carolina elementary schools participated in the multi-site Social and Character Development (SACD) Study, which evaluated programs designed to prevent aggression and promote positive social behavior.  We analyzed baseline (fall 2004) data about 691 third graders in 10 of the North Carolina schools.

We imputed missing data (11% of responses) using the expectation-maximization method.  We used SAS PROC FACTOR to conduct exploratory factor analysis (EFA) of 16 Likert-type items that teachers used to rate children’s behavior over the past month.  We used principal axis factoring, with squared multiple correlations as prior communality estimates.  Factor retention depended on the scree plot and the number of eigenvalues greater than 1.  We evaluated both orthogonal and oblique (promax) rotations.  We required a main loading of at least 0.60 and cross-loadings below 0.30 in order to retain an item.  We used an iterative process, deleting the item with the lowest main loading and/or highest cross-loading and then re-running the EFA, until we reached the final model.

We performed confirmatory factor analysis (CFA) using the lavaan package in the R environment.  Because the EFA and CFA used data from the same sample, both should be considered exploratory.  We tested both the final 3-factor model from the EFA and an alternative one-factor model.  To account for clustering of children within classrooms and schools, we adjusted fit statistics (root mean squared error of approximation [RMSEA], χ2/df, and comparative fit index [CFI]) using the Satorra-Bentler correction.  We evaluated concurrent and predictive validity by examining associations with demographic, behavior, and social information processing variables.

Results

The following results are preliminary.  The final EFA model included 10 items in an obliquely rotated solution with three factors (alpha=.84 to .89): inhibitory control, attention control, and impulsivity.  CFA indicated acceptable fit (χ²/df=5.18, RMSEA=.078 [90% CI .067, .088], CFI=.955).  As expected, effortful control was positively associated with female sex (F[1,689]=57.75, R²=.08) and negatively associated with aggression (r=-.65) and conduct problems (r=-.63).

Implications

Preliminary findings suggest that effortful control includes inhibitory control and attention control and is characterized by a lack of impulsivity.  Teacher reports measured all three constructs reliably; the effortful control scale appears to have sufficient reliability and validity to be used in subsequent analyses.  This effortful control scale includes content that is absent from others, suggesting that further development of an effortful control measure should utilize a larger and more comprehensive item pool.