Summary: The Self-Administered Interview© (SAI) serves to elicit eyewitness statements directly after the crime. Witnesses could still experience stress then. Because stress during retrieval produces memory-impairing effects, this study sought to compare the SAI with free recall under stress. An interaction between stress and interview was expected such that the SAI would elicit more comprehensive accounts than free recall in the control, but not in the stress group. One hundred and twenty-seven participants underwent a stress or control task. They witnessed a live staged crime and completed an SAI or a free recall. The SAI elicited a higher number of correct verifiable event details and a higher number of correct and incorrect perpetrator details than free recall. Accuracy rates were unaffected. Unexpectedly, despite causing moderate stress-induced cortisol elevations, stress exposure did not influence memory performance and did not interact with interview type. Hence, the SAI can safely be used, when witnesses are moderately stressed.Copyright © 2015 John Wiley & Sons, Ltd. When witnesses observe a crime, they may experience heightened stress levels. During the investigations, however, the police are likely to approach all witnesses to obtain eyewitness evidence, irrespective of their stress levels. A novel eyewitness interview procedure that is likely to be employed shortly after the incident when witnesses may still be stressed is the recently developed Self-Administered Interview© (Gabbert, Hope, & Fisher, 2009; see Hope, Gabbert, & Fisher, 2011, for a detailed description). To avoid problems of delayed interviewing (e.g., forgetting, Rubin & Wenzel, 1996, or misinformation effects, Gabbert, Memon, Allan, & Wright, 2004), this self-administered booklet is independently completed in writing by the witnesses at the crime scene. Its purpose is to secure an early witness statement when limited police resources do not allow conducting an immediate personal interview. The SAI is based on the cognitive interview (Fisher & Geiselman, 1992) and adopts some of its memory-enhancing techniques, such as mental context reinstatement, the report everything component, and multiple and varied retrieval. The tool hence provides ample retrieval support. The SAI is often compared with written free recall (FR) that provides only little retrieval support. Relative to FR, the SAI elicits a higher number of correct details—which usually amount to a large effect— but also a higher number of incorrect details. Importantly, however, accuracy rates (number of correct details reported divided by all reported details; Meissner, Sporer, & Susa, 2008) do not differ between the SAI and FR (Gabbert et al., 2009; Gawrylowicz, Memon, & Scoboria, 2014; Gawrylowicz, Memon, Scoboria, Hope, & Gabbert, 2014). That accuracy rates are unaffected by the SAI is essential because it shows that information gain does not compromise the amount of correct information that is elicited. In other words, the SAI increases information output relative to FR, while high accuracy rates are maintained. The SAI also pre