The treatment of emergencies in a hospital or in private practice is subject to a common set of rules, but each setting presents its own particularities. These specificities include the prevalence of different pathologies, the possibility of hospitalisation and the length of the observation period, and these are illustrated by pertinent case histories. The effect of different prevalence is well illustrated by the thrombolytic treatment of myocardial infarction. The cost/benefit ratio of this treatment is entirely different when administered to a population seen in private consultation, in which myocardial infraction makes up only 5% of patients presenting with thoracic pain, as opposed to a hospital setting in which the prevalence of myocardial infarction is much higher. The possibility of hospitalising patients is illustrated by community-based pneumonia, in which knowledge of the epidemiology and prognostic factors are guides in choosing appropriate antibiotics and in rapidly selecting the patients requiring admission to hospital. Finally, the length of observation is illustrated by patients presenting with abdominal pain. Certain symptoms and signs can indicate a surgical affection, but often a certain period of observation is required before the correct diagnosis can be made. To improve physicians' performance in emergency medicine will require better knowledge of areas up till now often ignored, such as epidemiology, decision analysis, a probabilistic approach to different pathologies, unfavourable prognostic factors of known illnesses, and simple clinical and paraclinical factors which serve to discriminate between those cases requiring hospitalisation and those which may be followed on ambulatory basis.