Within a prospective randomized screening study for early detection of colorectal cancer with rehydrated Hemoccult II test, the possibility of increasing the specificity of the test by retesting patients with an initially positive Hemoccult II test was investigated. Of those offered the test 3561 (62.6 per cent) returned it and it was positive in 210 cases (5.9 per cent). The repeat test was performed by 184 patients and was positive in 68 (1.9 per cent). All those with a positive initial test had rectosigmoidoscopy to 60 cm and a double contrast enema. A carcinoma was found in one in seven patients with a positive retest but in only one in 100 patients with a negative retest (P less than 0.001). The specificity of the test was, therefore, increased from 95 per cent to 98 per cent and the sensitivity was unchanged. Rescreening was offered at a later date and increased numbers were available: 7147 patients returned the test and 369 (5.2 per cent) were positive. The test was repeated in 360 patients and 118 (1.7 per cent) were positive. A colorectal neoplasm was found in one in three of those with a positive repeat test, compared with one in seven of those with a negative repeat test. In conclusion, screening for early detection of colorectal cancer with a rehydrated Hemoccult II test may be followed by investigation of only those patients with a positive retest. Such a procedure will reduce the work-load by 60 per cent without reducing sensitivity.