BACKGROUND: Demand oxygen delivery systems (DODS) allot oxygen by interrupting the oxygen flow during exhalation, when it would mostly be wasted. Because DODS conserve oxygen by various methods, there are important performance differences between DODS. We studied certain performance factors that have not previously been carefully examined. METHODS: A bench model was constructed to simulate a nose, airway, and alveolar chamber. A breathing simulator generated 4 respiratory patterns, at frequencies of 15, 20, 25, and 30 breaths/min. Eighteen models of DODS were tested at 4 settings, each up to the maximum output, and compared to continuous-flow oxygen. The variable of interest was the fraction of inspired oxygen (F(I)O(2)) in the alveolar chamber, which was measured for each condition. RESULTS: The DODS differed from continuous-flow oxygen, delivering 0.5-2.1 times (mean = 1.13 times) the F(I)O(2) increase at similar settings. During maximum output the DODS showed a wide range of F(I)O(2), from 0.27 to 0.46. There was a direct relationship between volume output per pulse in the first 0.6 s of inhalation and the delivered F(I)O(2). CONCLUSIONS: DODS settings were not equivalent to continuous-flow oxygen in a bench model assessment; with equivalent settings the DODS tended to deliver greater F(I)O(2) than did continuous-flow oxygen. The maximum output capacity differed markedly among the DODS, and the user should know the device's capacity. A volume-referenced setting system for DODS should be adopted that would allow more predictable oxygen prescription and delivery via DODS.