Aim. Different time frames have been used to ask about drinking in clinical and general populations. Surveys of clinical populations have asked about the quantity and frequency of drinking within the context of "when you were drinking". In general population surveys, the customary practice has been to ask about a period of time such as "the last 12 months". This paper compares answers to questions about drinking using both time frames. Design. Bivariate chi-squares and multi-variate logistic regression analyses were used to compare consumption estimates across the two time frames for different demographic and drinking categories of respondents. Setting and participants. In-person interviews were conducted with general population (N=3069) and representative treatment samples (N=381) in a northern California county. Measurements. Respondents were asked about their drinking within the context of "the last 12 months" and only "when you were drinking". Findings. There were no meaningful differences in aggregate measures of drinking based on the time frame of assessment in either sample. Drinking five or more drinks weekly was a significant predictor of consistent reporting of frequency of drinking among the clinical sample, and of reporting inconsistent frequency of drinking 12 or more drinks among the general population. Being female or being age 46 or over also was predictive of a "consistent" response in the general population for drinking 12 or more drinks; while being 46 or older, married, and white was predictive of consistent responses for drinking to intoxication in that population. Conclusions. Survey respondents do not average their drinking across a 12-month time frame that includes periods of abstinence; rather, they appear to answer only for the periods they were drinking.