Idiosyncratic or type B reactions are characterised by their unpredictability and lack of simple dose-dependency. They occur in a small proportion of patients, and usually the predisposing factors are unknown. A proportion of, but not all, idiosyncratic reactions are immune-mediated. Our understanding of immune-mediated reactions is based on the hapten hypothesis, which requires drug bioactivation, covalent binding to proteins, followed by uptake, antigen processing and a polyclonal immune response. The recently proposed 'danger hypothesis' can be considered to be additive to the hapten hypothesis. The hypothesis states that the immune system only responds when it detects danger. If no danger is detected, tolerance results. Thus, stimulation of an immune response to a drug-protein conjugate (signal 1) requires the presence of co-stimulatory signals and cytokines (signals 2 and 3), which propagate and determine the type of immune response. The nature of the danger signal is poorly defined, and has been proposed to include different forms of stress including chemical, physical and viral. Indeed, there are several examples where the frequency of drug hypersensitivity is increased in the presence of a viral infection, most notably in HIV disease. Nevertheless, this clinical evidence has to be regarded as being circumstantial and more direct experimental evidence is required to understand the role of 'danger' in the overall pathogenesis of drug hypersensitivity reactions.