Collaborative Strategic Reading (CSR) is a set of research-based strategies designed to improve reading comprehension, enhance students' content area learning, facilitate access to higher-level texts, and to promote student engagement. The present study examines how fidelity of implementation of CSR is associated with reading outcomes for students with mild to moderate disabilities using data from two nonoverlapping studies conducted in middle school language arts and reading classrooms (study 1) and social studies and science classrooms (study 2). Researchers use a definition of fidelity that includes both the amount of CSR instruction delivered by teachers and the quality of implementation. The following questions guided this research: (1) Is higher CSR instructional quality associated with increased student outcomes for adolescents in treatment classrooms, and for a subgroup of students with disabilities?; and (2) Is the amount of CSR instruction related to student outcomes for adolescents in treatment classrooms, and for a subgroup of students with disabilities? Researchers replicated the analysis with two different samples (studies 1 and 2), and compared results to examine the stability of estimated relationships, as well as to answer the secondary research question: How do the characteristics of study participants and program contexts influence the replication of findings? The first study took place in nine low income middle schools in the metropolitan areas of Denver, Colorado, and Austin, Texas. The second study was conducted in three middle schools located in the same large, urban school district in Colorado. Teacher and student participants were part of a school-district-and-university collaboration designed to evaluate the effectiveness of CSR within the context of a district initiative and scale-up study. The samples for both studies included the treatment group from larger randomized control trials of CSR. Teacher quality for both studies was assessed using the CSR Global Quality Rating Score measured through the "Implementation Validity Checklist." Teachers in both studies completed a log to monitor the dosage or amount of CSR intervention children received. The student outcome for both studies was reading comprehension ability, as measured by the reading comprehension subtest of the Gates-MacGinitie Reading Test (GMRT), Fourth Edition. Multilevel (random-intercept) models, as implemented in the software program Hierarchical Linear Modeling 7.0 were used to estimate the effects of CSR on outcomes controlling for pretest scores and student demographics. Both studies found that higher quality CSR instruction is associated with higher student outcomes for students in special education. The standardized coefficients in study 1, however, were nearly double those in study 2. Further research is needed to determine possible explanations for these findings. Tables are appended.