In 1996, 250 people living in the Dongting-Lake region of China were selected for a 2-year study. All had been or were infected with Schistosoma japonicum. All were treated with praziquantel, although eggs of S. japonicum were only detected in stool samples from 75 of the subjects. In 1998, 213 (85.5%) of these subjects, then with a mean (S.D.) age of 40.2 (14.2) years, provided stool samples for final assessment. Forty-nine (23%) of the 213 were found to be re-infected in 1998, with a geometric mean intensity of infection among the infected of 64.5 eggs/g faeces. The rate of re-infection was highest among those aged < 10 years, declining with increasing age, and higher in males than females (27.3% v. 8.3%; P < 0.005). The mean intensity of infection among the infected males was also higher than that among the infected females [72.4 (4.8) v. 17.8 (2.5) eggs/g; P < 0.005]. Water contact by the subjects was estimated from activity diaries, for 141 days over the 2-year study period, and expressed as skin exposure, in m2-min/day. The mean exposure of a group of subjects was calculated by detransforming the mean of the fourth-root-transformed (i.e. normalized) values for the exposures of each subject within the group. Overall, the 213 individuals had a mean exposure of 6.2 m2-min/day. Differences in occupation led to males having much higher mean water exposures than females (9.2 v. 1.1 m2-min/day). As there was an inverse association between age-specific exposure and age-specific re-infection intensity, the marked reduction seen in intensity of re-infection with increasing age is not attributable to decreasing exposure to water. Instead, the results of this 2-year cohort study provide evidence for age-dependent resistance to re-infection with S. japonicum. The 213 subjects who were followed up were classified, according to epidemiological outcome at the end of the study and the data on water contact, as 'susceptible' (N = 49; 23%), 'insusceptible' (N = 29; 13.6%) or of 'uncertain status' (N = 135; 63.4%). Thus, 78 subjects who are potentially informative in terms of immunogenetics were identified. Further investigation of these individuals should help to shed some light on the role of immunogenetic status in human immunity to Schistosoma japonicum.