The effects of program fidelity, gender, socioeconomic status, and school level were tested on various outcomes of a dating violence prevention program, Expect Respect. Fidelity data was collected from program facilitators, and individual posttest scores were gathered for individuals nested within each programmatic site. Multiple HLM models indicated that main effects for fidelity were present for the knowledge gained outcome scale, such that higher program fidelity led to higher posttest scores for participants. No other site level predictors affected outcomes or the fidelity-outcome relationship. Results from this study point towards the importance of implementing program fidelity when cognitive gains are a central goal of the program, whereas a more flexible program approach may be more optimal in conveying other programmatic components.