Monitoring the fidelity with which implementers implement behavioral procedures is important, but individuals tasked with monitoring fidelity may have received little training. As a result, their data may be affected by variations in implementers' performance, such as the frequency of fidelity errors. To determine the extent to which the frequency of fidelity errors affected data accuracy, we used a multielement design in which novice observers collected fidelity data from video models with three levels of programmed fidelity (40%, 80%, and 100%) of a procedure based on differential reinforcement of other behavior (DRO). Participants collected fidelity data with lower accuracy, as indexed by interobserver agreement scores, when errors were more frequent. These findings suggest that individuals collecting fidelity data may need additional supports to detect and record all errors when procedural fidelity is low.