SECRETS AND LIES: COMMENTS ON.

The article presents the author's views about a research done by researcher D.A. Dawson and his colleagues on alcoholism. The author says the good news is that if he had to point a student to a single large epidemiological study that was representative of the course of alcoholism in the United States, he could not point to a paper superior to that of Dawson's. The bad news is that Dawson and his colleagues are correct to warn us that their "cross-sectional data do not necessarily reflect the course of recovery across time for any individual." This acknowledgement makes many of their findings suspect. What comes out of computers, even if you add decimals, is only as good as what goes in. At the heart of the difficulties raised by their paper is a paradox. If you wish to predict an election, a large representative sample of the population, interviewed by a trained staff, is a necessity. If you wish to study secrets of which the owner does not wish to be aware, for example incest or not being in control of one's drinking over the past year, a small intensively studied, but often unrepresentative, sample whom the investigator has actually met, is a necessity. In some facets of alcohol abuse the first strategy is preferable, but if one wishes to study return to controlled drinking one needs to include other informants and to conduct observations over time until the "secret" emerges or can be ruled out on a case-by-case basis.