The aim of the present study was to investigate whether MRI could predict tumour involvement of the lateral resection margin on resected rectal cancer specimens. 26 specimens obtained after total mesorectal excision were examined in the sagittal and axial plane on a clinical 1.5 T MR system using phased-array surface coils. The shortest distance between the deepest point of tumour penetration and the lateral resection margin was measured. This distance is referred to as LRM. The results were correlated to measurements on digital images of giant histopathological sections. The total number of lymph nodes and lymph node metastases was also counted. LRM was less than or equal to 1 mm (a distance considered for non-radically excised tumours) in eight of the histopathological specimens. On MR images of the resected specimen, LRM was less than or equal to 1 mm in seven of these specimens. LRM was shorter in MR examinations than in histopathological giant sections in 11 specimens. The sensitivity, specificity, positive and negative predictive values for prediction of tumour involvement of the lateral resection margin (LRM < or = 1 mm) by MR of the resected specimen were 88%, 78%, 64% and 93%, respectively. Presence of lymph node metastases could not accurately be predicted by MR. However, it is concluded that the presence of a tumour free lateral resection margin can be predicted by MRI of resected rectal specimens when this exceeds 1 mm.