Eyewitness recall by 60 adolescents and 60 young adults in Experiment 1 and by 64 children and 63 preadolescents in Experiment 2 for a simulated theft in which gender-role characteristics and sex of criminal were manipulated (i.e., masculine male, feminine male, feminine female, masculine female) was investigated. Gender-role flexibility impacted on crime details and accuracy for criminal features and children's reports included fewer crime and criminal features and central crime details than did preadolescents. Children's gender-role beliefs differentially affected errors made when describing criminals with inconsistent gender-role characteristics. Adolescents provided fewer crime and criminal details and criminal features than did young adults, although both converted inconsistent into consistent gender-role information for the "feminine male" criminal. Forensic implications of the findings are discussed. (Contains 5 figures and 5 tables.)