Meta-analysis of social skills interventions of single-case research for individuals with autism spectrum disorders: results from three-level HLM.
Social-skills packages produce large, measurable gains for people with autism, especially when you track them with multiple-baseline or reversal designs.
01Research in Context
What this study did
Wang et al. (2013) pooled 115 single-case studies about social-skills training for people with autism.
They used three-level modeling to see how big the gains were and what made them bigger.
What they found
The average effect was large, but the design of the study mattered.
Multiple-baseline and reversal designs showed the strongest gains.
How this fits with other research
Kokina et al. (2010) and Karkhaneh et al. (2010) both looked only at Social Stories and saw small or unclear effects. Wang et al. (2013) included those papers and many others, so the big average effect comes from mixing lots of methods, not just stories.
Chan et al. (2021) later showed that physical-activity games can also give small-to-moderate social boosts, extending the toolkit beyond talk-based lessons.
Sung et al. (2019) moved the same idea into the workplace, finding positive gains for young adults with autism, so the large effects hold across ages and settings.
Why it matters
You can expect strong skill gains when you run social-skills packages, but pick multiple-baseline or reversal designs to prove it with data. Mix activities—games, role-play, job-practice—not just stories, to keep effects large and meaningful across clients.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Add a second tier or reversal phase to your current social-skills program so you can show clear, large effect on your graph.
02At a glance
03Original abstract
This meta-analysis used hierarchical linear modeling to examine 115 single-case studies with 343 participants that examined the effectiveness of social skills interventions for individuals with autism spectrum disorder (ASD). The average effect size of the included studies was 1.40 (SD = 0.43, 95% CL = 1.32-1.48, N = 115). In the further, several common predictors including intervention length, age and gender of the participants, and study quality indicators (provision of sufficient, in-depth, and replicable information of participants, settings/materials, independent variables, and dependent variables) were not found to mediate the intervention effectiveness. Only research design that the study employed was found to impact the intervention effectiveness; the studies using multiple baseline or reversal design had larger effect sizes than studies using other designs. Implications of the results and limitations of this study are discussed.
Journal of autism and developmental disorders, 2013 · doi:10.1007/s10803-012-1726-2