A critique of the official report on the evacuation of the World Trade Center: continued doubts.

This paper criticises the conclusions and the unanswered questions in the National Institute of Standards and Technology (NIST)'s official report on the evacuation of the World Trade Center in New York City, United States, on 11 September 2001. It reviews the extent to which the report disregards several conventional statistical methods and comments on the NIST's refusal to share the machine-readable data file with the scientific community for replication and further analysis. Problems lie in the sampling methods employed, the treatment of missing data, the use of ordinary least squares (OLS) with binary dependent variables, the failure to document the scalability of the scales used, the lack of tests to check for constant error variance, and the absence of overall fit tests of the model. There are also conceptual and theoretical issues, such as the absence in the report of considerations of the influence of group-level processes and their impact on the collective behaviour of evacuating collectivities.

Normative collective behavior in the Station building fire.

Objective. This article offers a test of the normative explanation of collective behavior by examining the fire at the Station nightclub in Rhode Island that killed 100 and injured nearly 200 persons.Methods. Information on all persons at the club comes from content analysis of documents from the Rhode Island Police Department, the Rhode Island Office of the Attorney General, and The Providence Journal. We use negative binomial regression to test hypotheses about the effects of group-level predictors of the counts of dead and injured in 179 groups at the nightclub.Results. Results indicate that group-level factors such as distance of group members at the start of the fire, the number of intimate relations among them, the extent to which they had visited the nightclub prior to the incident, and the average length of the evacuation route they used predict counts of injured and dead. The research also looks at what behavioral differences exist between survivors and victims, ascertains the existence of role extension among employees of the nightclub, and provides support for the affirmation that dangerous contexts negate the protective influence of intimate relations in groups.Conclusion. We argue for the abandonment of current emphasis on irrationality and herd-like imitative behavior in studies of evacuation from structural fires in buildings and for the inclusion of group-level processes in social psychological explanations of these incidents.