An epidemiological study on the G serotype distribution of group A rotaviruses (GARV) isolated in Kenya was carried out in one urban hospital in Nairobi and in two rural hospitals in Nanyuki and Kitui to clarify the prevalent G serotypes before future introduction of the ready licensed rotavirus vaccine in Kenya. A total of 1,431 stool specimens were collected from children, who were mainly outpatients, aged from 0 to 6 years old with acute gastroenteritis from August 1991 to July 1994. Samples positive for GARV by conventional ELISA were then analyzed by subgrouping and serotyping ELISA and by PAGE. To ascertain the G serotypes of viruses in samples that were unable to be typed by serotyping ELISA, polymerase chain reaction was also attempted. The prevalence of GARV was 28.4% in the urban hospital, 22.5% in Nanyuki, and 13.7% in Kitui. Among rotavirus-positive samples, subgroup II rotaviruses were detected in 63.1%, and subgroup I rotaviruses were 25.9%. Serotype G4 was most prevalent, accounting for 41.6% followed by 23.3% of serotype G1, 17.0% of serotype G2, and serotype G3 was rarely isolated. Seven strains of serotype G8/P1B rotavirus was detected for the first time in Kenya by RT-PCR. Eleven specimens with an unusual composition of subgroup, serotype, and electropherotype were atypical GARV in which the P-serotype was P1A, P1B, or P2. Although uncommon GARV serotype G8/P1B and atypical GARV were detected, the four major GARV serotypes, G1 through G4, should be targeted at this moment for vaccination to control this diarrheal disease in Kenya. Continuous monitoring of the G- and P-serotype distribution of GARV should provide important information about the impact of rotavirus vaccination in Kenya.