Most published studies of eyewitness testimony are flawed due to the types of samples used, the testing method, and the lack of attention given to the role of the subject's personal involvement in the situation being considered. The results of these studies may therefore not apply to real cases of abuse. The authors did several laboratory studies that tried to address some of these flaws. They included situations in which the children were participants rather than bystanders and used situations involving medical examinations, venipuncture, and inoculations so that the children would be reporting on experiences involving pain. Results did not support the view that stress interferes with a victim's memory. However, 3-year-olds had significantly less ability than 6-year-olds to respond correctly to objective and suggestive questions. Researchers studying children's testimony must simulate as many characteristics of actual events as possible if their studies are to be ecologically valid. Data tables and 75 references.