Violence exposure during childhood is common and associated with poor cognitive and academic functioning. However, little is known about how violence exposure influences cognitive processes that might contribute to these disparities, such as working memory, or their neural underpinnings, particularly for cognitive processes that occur in emotionally salient contexts. We address this gap in a sample of 54 participants aged 8 to 19 years (50% female), half with exposure to interpersonal violence. Participants completed a delayed match to sample task for emotional faces while undergoing functional magnetic resonance imaging scanning. Violence-exposed youth performed worse than controls on happy and neutral, but not angry, trials. In whole-brain analysis, violence-exposed youth had reduced activation in the left middle frontal gyrus and right intraparietal sulcus during encoding and the left superior temporal sulcus and temporal–parietal junction during retrieval compared to control youth. Reduced activation in the left middle frontal gyrus during encoding and the left superior temporal sulcus during retrieval mediated the association between violence exposure and task performance. Violence exposure influences the frontoparietal network that supports working memory as well as regions involved in facial processing during working memory for emotional stimuli. Reduced neural recruitment in these regions may explain atypical patterns of cognitive processing seen among violence-exposed youth, particularly within emotional contexts.

Recently in this journal, Weeland et al. (2018) published a thought-provoking article reporting moderating effects of children's serotonin transporter-linked polymorphisms (5-HTTLPR) on negative parenting during prevention with the Incredible Years series. Participants were parents and young children of 387 families enrolled in the Observational Randomized Control Trial of Childhood Differential Susceptibility study. An equally important finding, which we focus on in this comment, involved null effects for all tests of parenting as a mediator of prevention-induced improvements in children's externalizing behavior. Although such findings may seem surprising, both confirmations of and failures to confirm parenting change as a mediator of child behavior change are common in the prevention and intervention literatures. In this comment, we explore likely reasons for heterogeneity in findings, including both moderators of treatment effect size and methods used to test mediation. Common moderators of prevention and intervention response to Incredible Years include dose, parenting problems at intake, high-risk versus clinical nature of samples, how parenting is measured, and whether child training is included with parent training. All of these moderators affect power to detect mediation. We then discuss conceptual criteria for testing mediation in randomized clinical trials, and problems with interpreting mediating paths in cross-lag panel models. Although the gene effect reported by Weeland et al. is important, their cross-lag panel models do not provide strong tests of parenting as a mediator of child behavior change. We conclude with recommendations for testing mediation in randomized clinical trials.

In their commentary, Beauchaine and Slep (2018) raise important issues regarding research on behavioral parenting training (BPT). In this reply we highlight key points of agreement and respond to issues that we feel require clarification. BPT has been repeatedly proven effective in decreasing disruptive child behavior (also in the work of our research team). Yet, there is much to learn about for whom and how BPT is effective. Specifically, assessing the how (i.e., mediation) comes with many challenges. One of these challenges is taking into account the timeline of change, and being able to infer causal mechanisms of change. We argue that cross-lagged panel models (which we, and many other scholars, used) are a valid and valuable method for testing mediation. At the same time, our results raise important questions, specifically about the timing and form of expected changes in parenting and child behavior after BPT. For example, are these changes linear and gradual or do they happen more suddenly? To select the appropriate design, assessment tools, and statistical models to test mediation, we need to state detailed hypotheses on what changes when. An important next step might be to assess multiple putative mediators on different timescales, not only before and after, but specifically also during BPT.