This paper aims to assess and compare currently available antidotes for cyanide poisoning. Such evaluation, however, is difficult. Thus, extrapolation from the results of animal studies has potential pitfalls, as significant inter-species differences in response may exist, and these experiments often involve administration of toxin and antidote almost simultaneously, rather than incorporating a more realistic time delay before initiation of treatment. Direct inference from human case reports is also problematic; either because of uncertainties over the exposure levels involved (and hence the likely outcome without treatment), or because of difficulties in identifying the specific contribution of a particular antidote within the overall treatment regimen. Certainly an effort to compare the relative efficacy of cyanide antidotes produces equivocal findings, with no single regimen clearly standing out. Indeed, factors such as the risks of antidote toxicity to various individuals and other practical issues, may be more important considerations. There is therefore no single treatment regimen which is best for all situations. Besides individual risk factors for antidote toxicity, the nature of the exposure and hence its likely severity, the evolving clinical features and the number of persons involved and their proximity to hospital facilities, all need to be considered. Clinically mild poisoning may be treated by rest, oxygen and amyl nitrite. Intravenous antidotes are indicated for moderate poisoning. Where the diagnosis is uncertain, sodium thiosulphate may be the first choice. With severe poisoning, an additional agent is required. Given the various risks with methaemoglobin formers or with unselective use of kelocyanor, hydroxocobalamin may be preferred from a purely risk-benefit perspective. However the former alternatives will likely remain important.