ABSTRACT
Reading Recovery is one of the most researched literacy programs worldwide. Although there have been at least 4 quantitative reviews of its effectiveness, none have considered all rigorous group-comparison studies from all implementing nations from the late 1970s to 2015. Using a hierarchical linear modeling (HLM) v-known analysis, we examined if effects differed in the United States versus other nations, if experiments yielded larger effects than quasi-experiments, if the effects changed over time, and if the type of outcome mediated the impact estimates. We also considered the sustained effects of the intervention. After reviewing 203 primary studies, we identified 16 that met our criteria, such as treatment fidelity and experimental or high-quality quasi-experimental design. Based on a random effects model, the estimated overall effect was .59, with larger effects for outcomes based on the Observation Survey (Clay, 2013), and stronger effects in certain literacy domains, such as text reading, print knowledge, and general literacy. Although United States studies produced a larger point estimate (.61) compared to other countries (.52), and experiments (.69) yielded a larger estimate than quasi-experiments (.43), neither difference was statistically significant. Overall, effects did not change over time, but effects based on the Observation Survey did improve significantly from earlier to later studies. We also found that the long-term effect may diminish, but there were too few studies to estimate the sustained impact with confidence. The .59 overall effect places Reading Recovery in the top 10% in terms of impact of early literacy programs reviewed by the What Works Clearinghouse.
