The objective was to analyse clinical and non-clinical factors associated with the receipt of a prescription for a benzodiazepine among general practice patients. A survey of 110 consecutive patient encounters (consultations) as recorded by a representative sample of general practitioners in inner urban, outer urban and rural settings was designed. A total of 286 general practitioners took part during 1991-2. 31,256 patients (10,683 male; 34%) were surveyed and the odds of receiving a benzodiazepine script measured. Insomnia, unlike anxiety, was almost routinely managed with a benzodiazepine alone (insomnia 89.6%; anxiety 49.4%), whereas anxiety was more likely to be managed with non-drug management (insomnia 7.2%; anxiety 38.3%). In multiple logistic regression, the variables significantly associated with the prescription of a benzodiazepine included being a female patient, being an older patient and being an established patient, who attends a GP working in a busy practice in an inner urban area. A second regression model was run with the addition of three variables, namely the presenting problems of anxiety and insomnia, and the number of health problems. The only predictors of benzodiazepine prescribing in the full model were these three clinical variables together with patient age. There is a need to educate doctors about the non-drug management of insomnia. The stereotype of the doctor over-prescribing a benzodiazepine without an appropriate problem/diagnosis should be questioned. On the other hand, there is concern that patient age continues to be associated with a prescription of these medications, when all other clinical and non-clinical factors are taken into account.