Patterns of Skill Development in Preschool Cognitive Training Studies

• In Event: Unpacking the Black Box of Early Interventions through Broadened Measurement of Implementation Fidelity

Abstract

Computerized cognitive training paradigms that target domain-general skills—though once considered highly promising—are proving to be minimally efficacious (e.g., Aksayli et al., 2019; Shinaver et al., 2014). In contrast, there is some evidence that domain-specific mathematical training demonstrates some far transfer, albeit to related mathematical skills (e.g., Honoré & Noël, 2016; Park et al., 2016; Sella et al., 2016; Van Herwegen et al., 2018). However, in the case of both domain-general and domain-specific cognitive skill training, there has been little exploration of optimal training schedules. Most cognitive skill training studies with children use an intent-to-treat or treatment-on-the-treated approach to estimate effects of randomization to treatment condition. This assumes the training schedule implemented for the purposes of a given study is optimal or, alternatively, that there is a linear association between adherence to treatment and outcomes. However, it is likely that there are different rates at which children develop skills throughout training. While in the aggregate, participants in cognitive skill training tend to improve over time, some participants develop skills quickly and then their performance on training-related activities plateaus; others might develop skills more slowly and plateau later in training; others still might never see a plateau of their training-related performance because they complete too little training.
The goal of this study is to explore patterns of preschool-aged children’s training-related cognitive skill development. Using item-level data from three training conditions, we intend to pursue the following aims: (1) Test whether there are different patterns of training-related performance within and between two domain-specific mathematical skill and one domain-general training conditions, and (2) examine whether individual differences in baseline skills predict patterns of skill development in training. Results have the potential to inform how and for whom treatment is most effective by providing evidence for optimal schedules of training based on children’s baseline characteristics.
Recruitment is ongoing and data will be analyzed for approximately 70% of the final sample at the time of presentation. Participants are 324 four- to six-year-old children from approximately 26 income-restricted (Head Start and state-sponsored) classrooms. Children are randomly assigned (stratified by classroom) to one of three tablet-based training conditions (i.e., a non-symbolic number skills training condition, a symbolic number skills training condition, and an executive function training condition). Each child participates in training twice weekly for five weeks for a total of 10 sessions. Children complete a battery of assessments pre- and post-training. The battery includes tasks that evaluate several components of children’s domain-specific mathematical skills (non-symbolic number skills, symbolic number skills, standardized mathematics achievement) and domain-general skills (processing speed, executive function).
Latent growth mixture models—which estimate growth curves and capture individual variation around growth curves by estimating variances within each class—will be estimated to assess different patterns of training-related performance within and between training conditions. Once the appropriate number of classes is ascertained, predictors of membership in differing growth trajectories will be tested (whether baseline math, executive function, or processing speed predicts membership in a class). Implications for intervention design will be discussed.

Authors

• Andrew Ribner, University of Pittsburgh
  Presenting Author

• Lucia Huerga, University of Pittsburgh

• Melissa E Libertus, University of Pittsburgh