What Matters When Providing Conversational Recast Treatment? A Multilevel Modeling Analysis.

Abstract

Purpose: Conversational recasting treatment is generally effective. However, different versions of this treatment and different targets may yield different outcomes for children. Here, we directly compare multiple variations of conversational recasting to determine how modifications to delivery and target impact treatment outcomes.

Method: Using Bayesian hierarchical linear mixed modeling, we compared outcome data from 141 children with developmental language disorder (DLD) across 10 versions of recast treatment compared with enhanced conversational recast treatment (a version that incorporates linguistic variability and attentional cues). We also compared data from a subset of 132 children treated for one of five morphological targets to determine the relative difficulty of learning each target.

Results: Results suggest that seven of the 10 treatment conditions resulted in an average decrease in generalization probe performance relative to enhanced conversational recasting alone. These conditions resulted in probe performance that was between 47% and 84% worse than performance of children who received enhanced conversational recasting with no variations. The remaining three conditions were no better or worse than enhanced conversational recasting. One treatment target was easier to remediate, while the other four were no easier or harder to acquire through treatment.

Conclusions: These results suggest that different variations of conversational recasting facilitate or diminish learning by children with DLD and that clinicians' attention to the form and structure of treatment delivery is necessary to implement best practice for this treatment method. In contrast, clinicians may have more flexibility when selecting targets.

Supplemental material: https://doi.org/10.23641/asha.28055207.