Emergency Behaviour of Ferry Passengers Michael May Danish Maritime Institute, Hjortekaersvej 99, DK-2800 Lyngby, Denmark mim@dmi-online.dk Keywords: Emergency behaviour, Mustering simulation, Crowd psychology, Human Factors, Evacuation. Abstract Based on interviews and a questionnaire survey with 1200 passengers on Danish ferries, stipulations were made about the emergency behaviour of ferry passengers. In this part of the MEPdesign project (MEP = Mustering and Evacuation of Passengers) a special focus was given to initial delays expected in the mustering process, to potential way-finding problems caused by misinterpretation of signs, and to the effects of "group-binding" caused by the interest in finding relatives before being evacuated. The expected delays and counter-flows in the mustering process was later tested against a large scale mustering exercise with 600 passengers, but the results have to be evaluated and adjusted through the knowledge gained from real accidents as well as general knowledge about crowd behaviour in emergencies. Introduction This paper is about the methodological problem of obtaining information about the emergency behaviour of ferry passengers in the context of mustering and evacuation. For obvious reasons trained observers cannot plan to be present during real accidents. We therefore have to rely on theoretical assumptions and different forms of indirect evidence about passenger behaviour. In the MEPdesign project we had the unique opportunity to compare predictions about passenger behaviour and mustering time with a large-scale mustering exercise at sea with about 600 passengers. The exercise was made possible by Scandlines Denmark who, as a partner in the project, made the Ro-Ro ferry m/f Kronprins Frederik available for the exercise. It is perhaps tempting to understand a comparison between prediction and exercise as one between “theory” and “reality”, but this is where the major methodological problem arises from the Human Factors point of view. The reality of the exercise does not directly correspond to the reality of real accidents – or only to a certain subset of accident scenarios. Furthermore the theory from which the predictions were made, did also include empirical observations on walking speed, way-finding errors and effects of “group binding”, and these observations are valid independent of the data of the single exercise. This theory/reality issue is a first methodological problem. The theoretical assumptions and empirical data were used as input to the EVAC simulation model developed by Quasar and the purpose of the comparison between the predictions of the model and the data of the exercise, was also to evaluate the