People systematically misperceive social norms, overestimating peer support for inequitable beliefs. While correcting these misperceptions can catalyze social change, the mechanisms by which social learning facilitates belief revision remain poorly understood. In a preregistered experiment (N=903), U.S. participants estimated peer agreement with 12 gender norm statements across six domains, then received social information attributed to different sources: majority opinion (conformity cues), adults over 40 (age cues), highly educated individuals (prestige cues), or an arbitrary source (seasonal preference cues). All treatment conditions received identical social reference information derived from pilot data; only source attribution varied. We found that participants substantially overestimated peer support for inequitable norms (M=22.7 percentage points). Exposure to social information produced large reductions in misperception across all treatment conditions relative to control (Cohen's d = 0.42 -- 1.52), but effectiveness varied significantly by attributed source. Age and conformity cues produced the largest corrections (d=1.52, 1.48), significantly outperforming prestige cues (d=1.14) and arbitrary cues (d=1.16). These findings reveal that source attribution substantially moderates the effectiveness of norm correction interventions, even when informational content is identical, and suggest that prestige cues may be less effective for moral norm correction than for factual domains.