Researchers often use survey methods to study how people behave during building fire evacuations. However, reports on the accuracy of this information are lacking. In this study, individuals (n = 185) recalled details shortly after participating in evacuation experiments. This information was compared to the observed evacuation performances. Our findings suggest that the accuracy of self-reports on evacuation experiences is influenced by the type of information being recalled and reported. Results showed a perfect agreement for the recalled actions (Cohen's kappa k = 0.97) a moderate agreement for evacuation paths (k = 0.51) and time judgements (k = 0.47), a fair agreement for cue perceptions (k = 0.23) and a slight agreement for self-perceived roles (k = 0.04). While the results are more suggestive than conclusive, this study offers valuable insights into the potential biases associated with self-reported accounts in evacuation research.