Cross-examination permits styles of questioning that increase eyewitness error (e.g. leading questions). Previous research has shown that under cross-examination children change many of their initially accurate answers. An experiment is reported in which the effect of cross-examination on accuracy of adult eyewitness testimony was investigated. Twenty-two student witnesses watched a video of a staged theft, either in pairs, or individually. Paired witnesses discussed the video with their cowitnesses, but did not know they had seen slightly different versions. Participants in the co-witness condition demonstrated memory conformity and recalled less accurately than witnesses in the control condition. After approximately 4 weeks all participants were cross-examined by a trainee barrister. Following cross-examination there was no difference in accuracy between the two experimental groups. Witnesses in both conditions made many changes to their previous reports by altering both initially correct and incorrect answers. The results demonstrate negative effects of cross-examination on the accuracy of adult eyewitness testimony.