On August 19, 1955, a considerable portion of the city of Port Jervis, New York, was overwhelmed by floods. The water had barely receded from the streets when, in the middle of the night, a false report began to circulate to the effect that a tremendous dam above the city had broken under the weight of the flood waters. About a quarter of the city's inhabitants fled within an hour.

The following week a team of psychologists began an investigation of the incident. This report will present the findings from that investigation.

Three main considerations motivated the research project: (1) Disaster situations seem to provide an opportunity for the study of human interaction under conditions of stress. (2) The apparent similarity of natural catastrophes to those of wartime leads one to believe that predictive statements about population behaviors under natural disaster might hold for war disaster. (3) An examination of the pervasive and threatening rumor which stimulated the exodus from Port Jervis might reveal the structure of this kind of communication network.

It is clearly difficult, if not impossible, to reproduce the emotional components of disaster situations in laboratory experiments with human subjects. At the same time, there is evidence that simulated disaster situations, such as mock evacuations, do not produce essential aspects of catastrophes. For example, in Spokane, Washington (1) during a mock evacuation, the observed pace of residents reentering the city after the exercise was greater than their pace leaving the city. If research is to be done on disaster, it is necessary to conduct field studies of actual catastrophes in such a fashion that they provide comparable data, even though there is serious question about the kinds of hypotheses which can be tested under these relatively uncontrolled conditions.

Descriptive field studies in stricken communities indicate the kinds of problem which later can be examined more systematically. However, much of the data which have come out of these studies are not comparable in any meaningful way. While disjoint hypotheses can be offered from each, there is no way of deciding whether or not they apply to any other situation. The sampling techniques and the kinds of information which are collected vary markedly from study to study. In some instances, the reproducibility of data extracted from interviews is questionable. The suddenness with which disaster strikes, and the consequent need to carry out the field study quickly, create tremendous difficulties in planning a coherent piece of research.

In the absence of any unified theory from which an organized set of hypotheses might be drawn, the Port Jervis incident was examined largely in descriptive terms. A large body of data was gathered with the use of careful sampling, interviewing, and coding techniques, in the hope that it would be amenable to comparison to future studies, similarly conducted, and that it would begin to suggest hypotheses which could be tested and retested in other catastrophic situations.

Two kinds of data were collected. First, a series of interviews were conducted among officials in and around Port Jervis. A descriptive account of the rumor communication network and of behaviors of residents prior to and during the spread of the rumor was prepared from these data. Second, data were collected from a random sample of city residents, and (separately) from a saturation sample taken in the previously flooded section of the city. These data were analyzed to test some general hypotheses which were formulated before the field study was begun.

When we have once stated the unique characteristics of the disaster, we can begin to look into the kinds of stimuli which will affect the behaviors of the people involved. Even the limited comparisons which are available from previous field studies suggest that the effectiveness of measures taken to control activities in a disaster differs from situation to situation. Since an implicit applied problem in disaster research is the control of community behavior, it is of primary importance to be able to define the situation so that activities may be predicted and the proper measures for control can be instituted.

If the disaster studied is a “false alarm” or a rumor, as it was in Port Jervis, the question of the meaning of the data becomes even more complex. On an intuitive basis, one might say that while the false report is believed, reactions to it will not differ from reactions to a threat which turns out to be real. However, serious doubts are raised immediately as to whether post hoc reports given by respondents about their activities during the spread of a rumor will be valid to the same degree as their reports of an actual disaster. Specifically, at least two possibilities exist. (1) After a false report, the respondent may desire not to appear “taken in” and may color his description of the events accordingly, or (2) the respondent may be able to give a more realistic picture of the false report, because his fears are unrealized and he suffers less shock.

If the respondent is either more or less accurate in his report of a false disaster than in his report of an actual disaster, it will be relatively fruitless to try to compare interview data from the two kinds of situation. In the Port Jervis study, a careful estimate is made of the empirical validity of data obtained from the interviews; it remains to be seen whether a match exists with data from actual disasters.

If the match is poor, of course, the study of rumors will have to remain relatively independent from the study of real disaster. If the match is good, we can examine certain reactions to false reports without differentiating them from reactions to other catastrophes, until the period of denial. Further, we may compare the communications systems of disasters which differ largely with respect to whether or not their referents are real or imagined.

In order to facilitate the development of comparable data, the methodology of the study is presented in greater detail than usual (see especially Appendices A, Appendices B and Appendices C).