The author notes that survey methods could not be expected to capture completely the potential variability in the within- individual pattern of drinking, and for most purposes such temporal variation is ignored. Further, there is little evidence that relates population distributions of alcohol consumption over long periods to sample distributions typically based on much shorter time periods. There are two dimensions to the sampling exercise: the individual and the temporal. The discussion of alcohol-consumption data collection also considers the use of diaries in alcohol research on general populations, the use of high preset categories of quantity in the summarial approach, quota sampling, and probability sampling. In general remarks on data collection for alcohol, tobacco, and drug use, the author discusses the randomized response technique, differences in the methodology for the measurement of alcohol and tobacco use, and the need for more research on methods of collecting data on illicit drug use. 16 references