The grade of a tumour is gauged on the subjective assessment of a number of histopathological parameters. The problems associated with this exercise were viewed from a historical perspective and survival analysis of 447 patients receiving surgery for rectal adenocarcinoma was undertaken. Only deaths from rectal adenocarcinoma were included as events in the survival analysis. Seven grade-related parameters were scored by one observer. A grading system was constructed using the Cox regression model. The variables in the best-fitting parsimonious model comprised lymphocytic infiltration, tubule configuration and pattern of growth. Scores were derived from the model and a four grade system was created in which the groups were of similar size. Good reproducibility of the selected histopathological parameters was demonstrated. Grade-related parameters were then allowed to compete with stage-related parameters in an overall model of pathological prognostic categories. The parameters selected in the best model were number of affected lymph nodes, the presence of lymphocytic infiltration and extent of spread through bowel wall. A set of five prognostic categories was developed from this model.