While it is widely acknowledged by epidemiologists, pharmacologists and physicians that adverse drug reactions (ADRs) occur with considerable frequency, within the realm of medical sociology, drug-induced risk currently receives little critical attention. This paper looks into this medical scientific 'black box' to present a comprehensive account of the epistemological and political processes at play in the detection of ADRs. By focusing on the literature generated by pharmacologists, epidemiologists and others working in the field, this paper examines the various techniques and methods used to identify and calculate ADRs both during clinical drug trials and beyond. Although risk associated with drug consumption is often presented as a scientifically calculable objective phenomenon, the attribution of causal mechanisms in determining whether the drug has caused an adverse event is a highly contingent social process, often involving complex clinical judgements. During clinical trials, variables are controlled and exclusions are imposed in order to fulfil scientific protocol requirements. These exclusionary practices mean that major patient population groups such as women and the elderly are often underrepresented during the trial process. At the time a drug product license is granted many uncertainties exist about the risk of ADRs. Once a drug is in widespread use, a more comprehensive profile of risk may begin to emerge. However, given that in everyday use drugs can interact with other drugs, alcohol and even certain foods and that drug reactions can mimic the disease they are supposed to be treating, differentiating between 'signal' and 'noise' is a messy, contingent complex process.