Objective: To investigate a new definition of pain episodes among primary care low back pain consulters. Methods: Consecutive primary care low back pain consulters aged 30-59 years were mailed self-completion questionnaires. In the pilot phase [N = 259], two open-ended questions on episode inception were compared; four respondents were also interviewed in depth. In the main study, the reliability [N = 447] and validity [N = 189] of a question with discrete response categories were estimated. Results: Only 23 percent of pilot study participants gave comparable responses to the two episode inception questions. Interviews illustrated the difficulties patients found in answering the questions. Good reliability and validity of the question completed in the main study were indicated. Seventy-six percent of respondents gave consistent responses to the last pain-free month question at baseline and six months, and 79 percent of people gave responses at six-months that were consistent with their monthly questionnaires. Conclusions: Problems recalling the start of low back pain episodes indicate that wording of questions could lead to misclassification of patients. A question asking about the "last pain-free month" shows promise as a standard simple epidemiological measure of episode inception.