Eyewitness testimony plays a critical role in legal proceedings. Yet, its usefulness hinges on both the reliability of the information from the witness, and how that information is perceived. When eyewitnesses testify in a non-native language, language barriers may shape not only what is communicated but also how their testimony is interpreted and judged. This thesis investigates whether testifying in a non-native versus native language affects the reliability and perceived credibility of honestly reported eyewitness testimonies. Across three empirical studies, I examined how language influences memory accuracy, suggestibility, the confidence-accuracy relation, observer-rated credibility, investigative questioning practices, and non-verbal behavior. Study I comprised two experiments. In Study Ia (N = 121), mock eyewitnesses provided testimony in either a non-native (English; n = 61) or their native (Swedish; n = 60) language after viewing a mock-crime video. Testimonies accuracy and suggestibility were measured, and the confidence-accuracy relation computed. Witnesses also rated their own credibility and cognitive effort when giving the testimony. No differences emerged between witness groups in accuracy, suggestibility, cognitive effort, or self-rated credibility. However, in line with the confidence-accuracy relation, witnesses were more confident overall in correct than incorrect details. Importantly, non-native speakers reported significantly lower confidence than native speakers, despite being equally accurate. In Study Ib (N = 202), mock interrogators were randomly assigned to assess either a non-native or native video-recorded testimony for credibility and to formulate follow-up interrogation questions. The interrogators judged non-native speakers as less credible than native speakers. Study II examined the interrogation questions (N = 806) formulated in Study Ib with respect to how leading, doubt-casting, and open-ended they were. No differences emerged in the types of questions posed to non-native versus native speakers. However, when interrogators perceived a witness as more credible, they posed fewer questions casting doubt, but this effect was only observed for native-speaking witnesses. Study III examined whether the presence (vs. absence) of a set of predefined non-verbal cues predicted the accuracy of 680 statements, witnesses’ self-reported confidence, and observer-rated credibility. None of the cues reliably predicted these outcomes, challenging assumptions about the diagnostic value of non-verbal cues in legal contexts. Together, this thesis provides converging evidence that language barriers can introduce bias in how eyewitness testimony is received and acted upon, not due to differences in testimony reliability, but due to how the testimonies are perceived in terms of credibility. The findings underscore the need for legal systems to better account for challenges placed on non-native speakers and to adopt practices that promote more fair credibility assessments.