Three experiments examined fact-finders ability to judge the accuracy of verbal eyewitness testimony as a function of witness in-group/out-group status (based on ethnicity) and testimony presentation mode. In Experiment 1, Swedish fact-finders were presented with videotaped testimonies from high-and low-accurate in-group (Swedish) and out-group (immigrant) witnesses. Results revealed that fact-finders assigned more credibility to high- as compared to low-accurate in-group witnesses, but that actual witness accuracy was unrelated to credibility judgments of out-group witnesses. Path analyses showed that fact-finders accuracy judgments of in-group, but not of out-group witnesses were related to witnesses self-reported confidence. Fact-finders ability to discriminate between high-and low accurate in-group witnesses was found to generalize to written presentations of the testimonies in Experiment 2. Experiment 3 examined Swedish civilians, police officers, and judges ability to discern the accuracy of in-group and out-group witnesses answers to specific questions about a crime event as a function of testimony presentation mode (videotape or text transcript). Preliminary results indicate that police officers were more proficient than civilians and judges at discriminating between accurate and inaccurate statements from in-group but not of out-group witnesses. All three groups showed a strong response bias, in particular when judging statements of out-group witnesses presented in the visual modality. Implications for theory and legal practices are discussed.