AbstractA theoretical model is presented in which alcohol consumption patterns are expressed in terms of frequencies of initiation and probabilities of continued drinking once consumption has begun. The model is used to explicate the relationships between measures of drinking frequency, drinking quantity, modal or typical drinking and total alcohol consumption. A mathematical realization of this, model is developed and applied to data on quantities and frequencies of alcohol use obtained from a general population survey of California consumers. These data were used to estimate the basic parameters of the alcohol consumption model, obtain drinking pattern estimates derived from the model and analyse these measures in the context of background demographic variables. Pattern estimates from I fie model mere bench‐marked against, self‐reports available in the original data. The results of these analyses showed that: (1) frequencies of drinking were exponentially distributed; (2) probabilities of continued drinking were best characterized by a log‐logistic function; and (3) estimates of modal and total drinking levels derived from the model were substantively related to self reports. Estimates of total consumption obtained from the model were 28% greater than those obtained from standard quantity‐frequency estimates. This difference was consistent with expectations based upon the theoretica