[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address Please contact the publisher to request reinstatement.
[Skip to Content Landing]
Views 487
Citations 0
Comment & Response
July 29, 2019

Challenging the Association Between Screen Time and Cognitive Development—Reply

Author Affiliations
  • 1University of Waterloo, Waterloo, Ontario, Canada
  • 2University of Calgary, Calgary, Alberta, Canada
  • 3Alberta Children’s Hospital Research Institute, Calgary, Alberta, Canada
JAMA Pediatr. 2019;173(9):891. doi:10.1001/jamapediatrics.2019.2243

In Reply We appreciate the opportunity to engage in scholarly discussion surrounding our contribution to JAMA Pediatrics on screen time and children’s development.1 Ophir et al state that “all between-person associations were not significant.” This is incorrect because the confidence interval for the between-person association does not include zero. They also cite that the “within-person cross-lags were negligible.” We agree that these effect sizes, while significant, are small in magnitude. This is not surprising when one considers that the random-intercepts cross-lag panel model2 conservatively estimates the putatively causal influences of longitudinal associations by isolating the within-person directional association after extracting the temporal stability of constructs, as well as traitlike individual differences. Moreover, child development is multiply determined and single sources of influence rarely produce large effects. Ophir et al also suggest an analysis that contrasts complete cases vs those who have missing values. It is well documented that listwise deletion produces results that are more biased than statistical estimation procedures for handling missing data. Accordingly, we used full information maximum likelihood estimation.3 In addition, Ophir et al request a statistical comparison of the direction of cross-lags (screens to milestones vs milestones to screens). As they note, the overlapping confidence intervals indicate that cross-lags are not significantly different from one another. We agree with this point and merely attest that the association from screens-to-milestones observed in our data are not zero. Ophir et al also discuss confounding variables. We did not have measurement on confounds at the same time-series schedule, prohibiting the modeling of dynamic effects for these factors over time. However, the application of the RI-CLPM does rule out between-person confounding statistically.2 We encourage other researchers to consider time-varying mediators and moderators of the screen time–child development relationship. Finally, and perhaps most notably, Ophir et al suggest that the “moral panic of screens is not justified.” We have drawn carefully hedged conclusions based on patterns of association. While the media may have portrayed our findings as indicative of panic, we have not suggested as much. We believe that screen time is one factor, among many, that plays a role in children’s development.