Data were obtained from RCMP files in a Vancouver suburb. All robberies committed between 1987 and 1989 and all fraud cases committed in 1989 were examined. Eyewitness descriptions of perpetrators were recorded, and the level of detail provided by eyewitnesses was noted. Of 170 perpetrator identification attempts, 154 (90.58 percent) were with photo spreads. Analysis indicated that fraud typically took longer to reach the identification stage than robbery, but more fraud cases involved an identification attempt. Within robbery cases, many factors varied, such as presence or absence of a weapon, presence or absence of witnesses, and relationships between witnesses and victims. Within fraud cases, the nature of interaction between victims and perpetrators varied. The forensic context also revealed natural confounds such as the victim's closer proximity to the perpetrator and subsequent higher arousal. Overall, results demonstrate that differences between and within crime types can affect eyewitness recall and recognition. Consequently, researchers of eyewitness memory need to be cautious about generalizing results from one context to another. 29 references, 3 notes, and 5 tables