First graders (N = 323) participated in an evidence-based peer-mediated classwide reading program (Peer-Assisted Learning Strategies; PALS). A "dual-discrepancy" approach was used to identify 66 children unresponsive to PALS. Unresponsiveness was defined as performance levels and growth rates substantially below those of average readers based on Curriculum-Based Measures. An exploration of this approach revealed that the dual-discrepancy approach reliably distinguished among unresponsive at-risk, responsive at-risk, and average-performing readers. Unresponsive students were assigned randomly to one of three increasingly individualized treatments: PALS, Modified PALS, or one-to-one tutoring by an adult. The relative effectiveness of the three treatments was evaluated by comparing the three groups' performance on phonological awareness and reading-related measures. No statistically significant between-group differences were found. Effect sizes comparing the treatments and proportions of nonresponders following treatment suggest that one-to-one tutoring was the most promising for reducing unresponsiveness. Implications for further research and service delivery are discussed.