OBJECTIVE: Male circumcision reduces risk of HIV among heterosexual men by about 60%. Modelling the impact of circumcision on HIV transmission, and planning service expansion, relies on self-reported circumcision status. We investigated the validity of self-reported status. METHODS: Survey and in-depth interview (IDI) data from adolescents enrolled in a community randomized sexual health intervention trial in rural Mwanza, Tanzania were analysed. RESULTS: The 5354 male school attenders (median age 15.5 years) were recruited in 1998 and followed for 3 years. At baseline, circumcision prevalence was 13.7% by self-report and 11.8% by clinical examination, rising to 17.3% by clinical examination at final survey. Only 61.5% of Muslim males were circumcised at the final survey. Of 506 participants who reported being circumcised at baseline, only 78.9% reported this at interim. Similarly, only 84.2% of participants clinically assessed as circumcised at baseline were also assessed as circumcised at interim. At both baseline and interim surveys, about 80% of participants who reported being circumcised were also found to be so at clinical examination. There was a high tolerance and respect for circumcision among male IDI respondents, with widespread belief that it was beneficial for penile hygiene and disease prevention. The majority of female IDI respondents said that they did not know what male circumcision was. DISCUSSION: Attitudes to male circumcision were positive in this population despite its low prevalence. There were substantial inconsistencies in both self-reported and clinically assessed circumcision status. Methods are needed to improve self-report and training of clinicians in this setting.