Methods for classifying chronic pain in population studies are highly variable, and prevalence estimates ranges from 11% to 64%. Limited knowledge about the persistence of pain and the validity of recall questions defining chronic pain make findings difficult to interpret and compare. The primary aim of the current study was to characterize the persistence of pain in the general population and to validate recall measures against longitudinal reporting of pain. A random sample of 6419 participants from a population study (the HUNT 3 study in Norway) was invited to report pain on the SF-8 verbal pain rating scale every 3 months over a 12-month period and to report pain lasting more than 6 months at 12-month follow-up. Complete data were obtained from 3364 participants. Pain reporting was highly stable (intraclass correlation 0.66, 95% confidence interval 0.65 to 0.67), and the prevalence of chronic pain varied considerably according to level of severity and persistence: 31% reported mild pain or more, whereas 2% reported severe pain on 4 of 4 consecutive measurements. When defined as moderate pain or more on at least 3 of 4 consecutive measurements, the prevalence was 26%. Compared with the longitudinal classification, a cross-sectional measure of moderate pain or more during the last week on the SF-8 scale presented a sensitivity of 82% and a specificity of 84%, and a sensitivity of 80% and a specificity of 90% when combined with a 6-month recall question. Thus pain reporting in the general population is stable and cross-sectional measures may give valid prevalence estimates of chronic pain. Copyright © 2012 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.