Symptom duration is integral to clinical and epidemiological research on pain. It is widely used for sample selection and commonly assessed in clinical practice. However, there has been little specific investigation of the link between duration and outcome. This work aimed to examine the association of episode duration with symptoms and clinical course in primary care consulters with low back pain (LBP). In a cohort of general practice LBP consulters, 619 patients returned two or more questionnaires during a 12-month period. LBP episode duration at baseline consultation was defined by time since their last pain-free month. Associations of duration with pain, disability and psychological status at baseline were examined. Survival analysis was used to assess the association between episode duration and time taken for disability to improve by 30%. At baseline, increasing episode duration was associated with worse pain, disability and psychological status (p < 0.001), but there were no differences between people with more or less than 3 months of pain. People with 3 years' or more duration at baseline took significantly longer to improve than those with shorter duration (adjusted hazard ratio 1.57, 95% confidence interval 1.27-1.95). In conclusion, memory of LBP episode duration is associated with pain, disability and psychological status, and is an independent predictor of time to improvement. There are important differences between people who recall more or less than 3 years' duration. Mechanisms for these associations are poorly understood, but this research suggests that duration itself is an important focus for research.