Three studies investigated whether requiring eyewitnesses to provide multiple accounts of a crime prompted them to exhibit more accurate and complete recall (i.e. hypermnesia) or towards providing more inaccurate and confabulated testimony. Subject-witnesses viewed videotapes enacting several types of crimes. Subjects exhibited hypermnesia in two studies in which they were asked to provide accounts of the incident via a free-recall procedure. Indeed, there tended to be roughly a 10–20 per cent increase in the number of facts accurately recalled from the initial interview to the third. The number of incorrect recollections and confabulations did not reliably increase. In the second experiment, hypermnesia was observed even when eyewitness memory was degraded by delaying the initial interview until a week after the witness viewed the crime. In a third study, which employed a multiple-choice questioning procedure, subjects exhibited neither hypermnesia nor heightened incorrect responding. Discussion centres on possible factors influencing the hypermnesia effect and implications of the present research for theory on hypermnesia.