A study was conducted to examine the false recognition effect in criminal profiling. Participants (N=83) were presented with a crime scene description and a criminal profile made up of semantically related statements. After a one week delay, the task for each participant was to recognize statements from a suspect description that were included in the profile, either with or without the profile in front of them. The suspect description consisted of descriptors that were: (1) included in the profile, (2) related to, but not included in the profile, and (3) unrelated to the profile. Participants in the memory-based condition also had to indicate, for each recognized item, whether they actually remembered reading the statement or simply knew that they had read it. Results indicate that participants who had to rely on their memory were more likely to recognize descriptors that were related to the profile, but not included in it, and these participants frequently indicated that they remembered, rather than simply knew, that they had experienced these statements. Potential explanations for these findings, implications for profiling practices, and directions for future research are discussed.