Since their introduction in 1985, oxygen concentrators have only been recommended when domiciliary oxygen is used for over 8 h day-1. Subsequent changes in the prices of oxygen merit a reappraisal of the prescribing of concentrators and cylinders when oxygen is used for less than 8 h day-1. Twenty-six patients in two health districts who used oxygen for less than 8 h day-1 completed a crossover study in which each group received oxygen from each source for consecutive 3-month periods. The patients were visited at home before and during the study, and on each visit they completed a questionnaire asking about their use of oxygen, how acceptable they found the two sources and about several dimensions of their quality of life. The theoretical minimum cost of cylinder supply, the actual cost of cylinder supply and the average concentrator costs were assessed. The patients found the concentrators to be more acceptable, more useful and less obtrusive than cylinders. They used more oxygen in more rooms of the home during treatment with concentrators, and there were improvements in the quality-of-life measurements. The costing information showed that, both in theory and in practice, oxygen concentrators are cheaper than cylinders when oxygen is used for more than about 1.4 h day-1. These results suggest that the provisions for the supply of domiciliary oxygen should be reviewed and that concentrators should be recommended for patients who use more than around 1.4 h day-1.