Evidence‐gathering begins at the scene of an incident. Available witnesses may be asked to provide an account of what happened, either in response to an open request for information or, in some regions, by completing a Self‐Administered Interview (SAI©). In both cases, an investigative interview may be conducted at some later date. This study sought to determine the impact of an initial retrieval attempt on a subsequent interview. After exposure to a mock crime, participants completed an SAI© or a free recall (FR), or did not engage in an initial retrieval (Control). One week later, participants were interviewed with a Cognitive Interview. SAI© participants reported more correct information and maintained higher accuracy than FR and Control participants. Consistency analyses revealed that the SAI© was effective because it preserved more of the originally recalled items (Time 1) than did an initial FR, and not because it yielded new recollections at Time 2. Copyright © 2014 John Wiley & Sons, Ltd.