Objective: One of the most difficult challenges confronting alcohol researchers is the assessment of specific risks associated with specific drinking levels (one, two or three or more drinks). In this article, a model-based method is presented that provides specific measures of dose, exposures and risks related to drinking and enables adequate assessment of dose-response relationships using survey data. Method: Survey data collected from 2,102 college students on drinking patterns and related problems were analyzed using a mathematical model that extracts separable measures of dose (number of drinks consumed per occasion) and exposures (numbers of occasions) for every respondent. A constrained class of quadratic risk models was estimated that relates units of exposure at each dose (e.g., number of occasions drinking five drinks) to problem outcomes (e.g., number of hangovers). The resulting dose-response functions provided an estimate of the rate of problems produced on each occasion of drinking at a given drinking level. Population risks were estimated by reintegrating dose response with the population distribution of drinking exposures. Results: The distribution of drinking exposures among college drinkers was markedly peaked at a modal dose of two drinks, with variations in light and heavy drinking between demographic groups. Multivariate analyses showed relative rates of heavier drinking were greatest among white male freshmen and unrelated to residential settings. Most drinking problems showed a marked dose-response relationship to drinking levels. Population risks, however, were typically unimodal, with many problems reported at moderate drinking levels (two or three drinks). Conclusions: Distributed across a large number of drinkers and drinking occasions, problems reported by students occurred most often at moderate drinking levels (two to three drinks). Nevertheless, heavier drinking remained substantively related to many drinking problems. These observations encourage aview of college drinking that supports interventions related to both moderate and heavy drinking.