A retrospective study of notified hepatitis B virus (HBV) infection in Edinburgh during 1975–92 identified 525 acute cases. For 343 where a probable transmission route could be determined, 215 were due to shared equipment by injection drug users (IDUs), 29 to homosexual intercourse, 25 to heterosexual or household contact with IDUs, 21 to heterosexual contact with infected non-IDU partners and 53 to various other or multiple routes. Cases were unevenly distributed geographically, particularly those among IDUs. The highest incidence within a post code district was approximately 2·5 times that for all Edinburgh. Annual cases peaked in 1984 then declined to low levels in the early 1990s. This reduction was most marked among IDUs, and may be ascribed both to changed injecting behaviour and decreased susceptibility within this group. The latter factor implies that HBV infections may be an unreliable guide to human immunodeficiency virus (HIV) infection in populations where HBV is highly prevalent.