Psychiatric diagnosis depends, centrally, on the transmission of patients’ knowledge of their experiences and symptoms to clinicians by testimony. In the case of non-native speakers, the need for linguistic interpretation raises significant practical problems. But determining the best practical approach depends on determining the best underlying model of both testimony and knowledge itself. Internalist models of knowledge have been influential since Descartes. But they cannot account for testimony. Since knowledge by testimony is possible, and forms the basis of psychiatric diagnosis, its very existence is a factor in support of an externalist model of knowledge in general. Internalist and externalist models of knowledge also suggest different ways of responding to the practical challenges of basing psychiatric diagnosis on testimony. Thus the argument in favour of externalism also supports a potentially empirically testable hypothesis about interpretation of non-native speakers for accurate psychiatric diagnosis: interpretation of non-English speakers should be as transparent and unhindered by specialised medical knowledge as possible.