The current paper examines critically the literature on deaths attributed to heroin overdose, and examines the characteristics and circumstances of such deaths. In particular, the dominance of the widely held belief that heroin-related fatalities are a consequence of overdose is challenged. Deaths attributed to overdose represented in the literature are typically older, heroin-dependent males not in drug treatment at the time of death. Fatalities involving only heroin appear to form a minority of overdose occasions, the presence of other drugs (primarily central nervous system depressants such as alcohol and benzodiazepines) being commonly detected at autopsy. Furthermore, deaths attributed to overdose are likely to have morphine levels no higher than those who survive, or heroin users who die from other causes. It is concluded that the term overdose is, in many cases, a misleading term, since it implies the same mechanism of death in all cases, an implication that is neither clinically useful nor consistent with published data. Implications for the prevention of heroin-related deaths are discussed.