The extent of tumour growth beyond the muscularis propria (mesorectal spread) was measured in specimens from 167 consecutive patients with rectal cancer. The 5-year survival was significantly greater in patients with slight mesorectal spread (4 mm or less) than in those with more extensive mesorectal spread (55% [95% confidence interval 42-66%] vs 25% [13-38%]). The prognostic value for survival of mesorectal spread was independent of the presence of lymphnode metastases. There were also significant differences in survival between patients with slight and extensive mesorectal spread among patients with Dukes' stage B tumours (66% [41-82%] vs 37% [14-60%]) and those with Dukes' stage C tumours (30% [12-52%] vs 18% [6-34%]). Thus mesorectal spread of rectal cancer is an important determinant of survival, and its accurate measurement may serve to subdivide Dukes' B and C cases. In this study tumour involvement of the lateral resection margin was not a useful predictor of local recurrence, but it did correlate with poor prognosis.