Eyewitnesses sometimes recall things at later interviews that they did not recall at previous interviews (reminiscence). When these cases are argued in the courtroom, attorneys may claim (and judges may warn jurors) that eyewitnesses who provide reminiscences are necessarily inaccurate witnesses. Consequently, their testimony may be prematurely discredited or dismissed. We examined here the role of varying the retrieval cues across interviews to account for reminiscence. Participants watched a videotaped mock crime and were tested for recall on two occasions using the same or different cues. Results supported the hypothesis that varying retrieval cues increases the amount of reminiscence. Furthermore, nearly all participants exhibited some reminiscence. Finally, reminiscence was not significantly correlated with overall accuracy of testimony. These findings suggest that many of the assumptions underlying legal tactics and judges' instructions regarding reminiscent inconsistencies are erroneous. Copyright © 2006 John Wiley & Sons, Ltd.