Dukes' staging is the most common means of staging and grouping colorectal carcinomas and is also used to determine which patients will be offered adjuvant therapies or entered into clinical trials. This study was performed to assess the degree of variation in the staging of colorectal carcinomas in normal clinical practice. Seven consultant surgeons and two consultant pathologists returned questionnaires asking them to stage 14 carcinomas on the basis of their pathology reports alone. The results show agreement among all nine in only six out of the 14 cases. In two cases there was a close to 50:50 split in perceived stage. Between them, the nine consultants produced eight different sets of staging results. These results indicate difficulties in the application of Dukes' staging system for several possible reasons. There may be misinterpretation of the written report, misapplication of the staging system because of unfamiliarity or confusion between the various modifications of Dukes' system which have been published. A more precisely defined staging system based on a standard proforma may be more appropriate in modern clinical practice.