A previous study has shown a lower survival for colorectal cancer in Europe than in the United States of America (USA). It is of interest to examine the extent to which anatomical location and morphological type influence this difference in colorectal cancer survival. We analysed survival for 151,244 European and 53,884 US patients diagnosed with colorectal cancer aged 15-99 years during the period of 1985-1989, obtained from 40 cancer registries that contribute to the EUROCARE study from 17 countries, and nine Surveillance, Epidemiology and End-Results (SEER) registries in the USA. Cases included in the analysis were first primary malignant tumours (ICD-O behaviour code 3 or higher). Relative survival was estimated to correct for competing causes of mortality. The Hakulinen-Tenkanen multiple regression approach was used to examine the prognostic impact of sub-site and ICD-O histology codes. Relative excess risks (RERs) derived from this approach estimate the extent to which the hazard of death differs from that in a reference region after adjustment for mortality in the general population. In order to explore geographical variation, we defined three groups of European registries within which survival rates were known to be broadly similar. The proportion of cases with unspecified sub-site was higher in Europe than the USA (10% versus 2%), but sub-site distributions were broadly similar in the two populations. With the exception of appendix, 5-year survival was 13-22% higher in the USA than in Europe for each anatomical sub-site. The proportion of non-microscopically-verified cases was higher in Europe than the USA (16 versus 3%). Adenocarcinomas arising in a polyp (ICD-O-2 8210, 8261, 8263) were more frequent in the USA than Europe (13 versus 2%). Five-year survival was higher in the USA than Europe for each morphological group, with the exception of non-microscopically-verified cases. When age, gender and sub-site were considered, RERs ranged from 1.52 to 2.40 for the European populations (with the USA as a reference). After inclusion of morphology codes, the range of RERs fell to between 1.28 and 1.86, mainly because of the high frequency of adenocarcinoma in polyps in the USA. This analysis suggests that the large survival advantage for colorectal cancer patients in the USA can only marginally be explained by differences in the distribution of sub-site and morphology. The main explanatory difference is the proportion of adenocarcinoma in polyps.