AIMS: To study the expression of Ki-67 and cytokeratin 20 (CK20) in a group of hyperplastic polyps (including a group with "atypical" features) with the aim of determining whether upper crypt Ki-67 staining and lower crypt CK20 staining correlated with these atypical features, as assessed by light microscopy. METHODS: Fifty seven formalin fixed, paraffin wax embedded hyperplastic colorectal polyps from 53 patients were selected on histological grounds; these comprised 26 typical polyps and 31 with atypical features, which included nuclear hyperchromatism, basal crowding, and increased mitotic activity. These polyps were examined using a standard immunohistochemical method with antibodies against CK20 and Ki-67. Comparisons were made with normal mucosa, adenomatous polyps, and carcinomas. RESULTS: Of the 26 typical polyps, 17 showed the usual pattern of lower crypt Ki-67 and upper crypt CK20 staining; one with upper crypt Ki-67 staining but normal surface CK20 staining; seven with Ki-67 confined to the lower half of crypts but with scattered lower crypt CK20; and one with both upper crypt Ki-67 staining, together with scattered CK20 basal staining. Of the 31 polyps with atypical features, 11 showed the usual staining pattern of lower crypt Ki-67 staining and surface staining with CK20; two showed Ki-67 staining extending into the upper half of crypts, but with a normal surface staining with CK20; 14 showed Ki-67 confined to the lower half of crypts, but scattered lower crypt staining with CK20; and four showed upper crypt Ki-67 staining together with scattered CK20 lower crypt staining. CONCLUSIONS: The normal pattern of lower crypt Ki-67 and upper crypt CK20 was seen in 28 of the 57 hyperplastic polyps and, in general, this corresponded with standard light microscopic appearances. Twenty one of the 57 polyps showed lower crypt mosaic CK20 staining, which in general corresponded with basal abnormalities on light microscopy, although seven specimens had normal appearances. Two smaller subsets emerged, one showing upper crypt Ki-67 staining in the presence of normal CK20 expression (three cases) and another in which a combination of lower crypt CK20 and upper crypt Ki-67 expression was seen (five cases). This last pattern was similar to that of neoplastic polyps and raises the possibility that a subgroup of hyperplastic polyps exists that may be a variant with malignant potential. Further studies with markers of mismatch repair genes and K-ras mutations may help to clarify this issue.