The relationship between symptom duration and long-term survival following colorectal cancer is complex, and a number of factors may influence the length of time from onset of symptoms of cancer diagnosis. We prospectively studied 777 consecutive colorectal cancer patients to determine the association between symptom duration and survival independent of other clinical and pathological features. We used survival curves, the logrank test and Cox's proportional hazards model to assess possible changes in relative risk of death with increasing symptom duration, without making any a priori assumptions. We found that symptom duration shortened with advanced tumour stage (P < 0.0006) and was also shorter for patients presenting with bowel obstruction (P < 0.0001). Univariate survival analysis showed that long-term survival increased consistently with symptom duration (P < 0.001). However, when the effect of tumour stage and bowel obstruction were accounted for in a multivariate analysis, no decrease in the relative risk of death was seen as symptom duration increased. The addition of other variables to the proportional hazards model such as age, sex or tumour site did not further influence the risk function form of symptom duration. Our results suggest that early diagnosis of colorectal cancer should remain our goal when assessing patients with suggestive gastrointestinal symptoms.