To investigate the relationship between survival in colorectal cancer patients and the number of lymph nodes examined by a pathologist, previously attributed to stage migration, we used data from a cohort of 5174 colorectal cancer patients recruited between September 1991 and August 1994, and followed-up for 5 years. We selected cases with data present on all prognostic variables, and stratified them into three groups by number of nodes examined. We made a multivariate survival comparison using a Cox regression model. In all, there were 3592 cases with data present on all prognostic variables. Patients who had> 10 nodes identified had a significant survival advantage over those who had 5–10 identified, who had in turn a similar advantage over those with 0–4 identified (P< 0.001). This effect was present in the whole group and at all Dukes' stages, although statistically significant only in stages B (P= 0.004) and C (P= 0.019). The effect remained after adjustment in a Cox regression model in which the mean number of nodes taken out by each surgical firm did not predict survival. In a sub-group with data on lymphocytic infiltration into the primary tumour a survival advantage was noted in those with prominent rather than mild infiltration (P< 0.001): the former also tended to have more nodes found (P= 0.015). Stage migration alone cannot explain these results, as survival advantages are noted across the whole population independent of stage. Lymphocytic infiltration into the primary tumour is prognostically important, and is associated with the number of nodes found. Reactive enlargement of lymph nodes in the mesentery may make them easier to find, reflect immune response to the tumour, and thus indirectly impact upon survival.