Caffeine in drinking water was offered ad lib. to male and female Syrian golden hamsters (Mesocricetus auratus W) for 90 days. Animals were randomly assigned to three dose groups (91.3, 274 and 822 mg/litre) and one control group (filtered tap water), each consisting of 20 male and 20 female animals. In relation to body weight, mean caffeine consumption was higher in females (low dose: 14.7; medium dose: 50.8; high dose: 104.8 mg/kg body weight/day) than males (low dose: 9.0; medium dose: 24.6; high dose: 65.2 mg/kg body weight/day). Caffeine in plasma was measured after 3 days, 3 wk and 3 months of treatment. As expected from calculation of the caffeine intake, mean values were higher in females (low dose: 0.6; medium dose: 3.6; high dose: 7.2 mg/litre) than in males (low dose: 0.4; medium dose: 0.7; high dose: 2.9 mg/litre). After 3 days of treatment, a transient, non-dose-related increase in mean (SEM) tri-iodothyronine (n=10) was found in the medium and high-dose (751+/-23 and 742+/-25 ng/litre, respectively) groups of males compared with that in the controls (610+/-39 ng/litre)(P<0.05). The values measured at later time points (days 24 and 91) were similar in all groups. No treatment-related changes were found in thyroxine (days 3, 24 and 91) and other clinicochemical analytes (day 91), absolute and relative adrenal weight (day 91), gross pathology and thyroid histopathology (day 91). In conclusion, no signs of thyroid toxicity of caffeine were observed in the Syrian golden hamster.