Infusing nitric oxide at a constant rate into a breathing circuit with intermittent mainstream flow causes formation of nitric oxide pools between successive breaths. We hypothesized that incomplete mixing of these pools can confound estimates of delivered nitric oxide concentrations. Nitric oxide flowed at a constant rate into the upstream end of a standard adult breathing circuit connected to a lung model. One-milliliter gas samples were obtained from various sites within the breathing system and during various phases of the breathing cycle. These samples were aspirated periodically by a microprocessor controlled apparatus and analyzed using an electrochemical sensor. The pools of nitric oxide distorted into hollow parabolic cone shapes and remained unmixed during their propagation into the lungs. In our preparation, time-averaged nitric oxide concentrations were minimal 60 cm downstream of the infusion site (18 ppm) and maximal 15 cm upstream of the Y-piece (36 ppm). The concentrations were mid-range within the lung (23 ppm), yet were substantially less than predicted by assuming homogeneity of the gases (31 ppm). Generally, nitric oxide concentrations within the lung were different from all other sites tested. Incomplete mixing of nitric oxide confounds estimates of delivered nitric oxide concentrations. When nitric oxide is infused at a constant rate into a breathing circuit, we doubt that any sampling site outside the patient's lungs can reliably predict delivered nitric oxide concentrations. Strategies to ensure complete mixing and representative sampling of nitric oxide should be considered carefully when designing nitric oxide delivery systems.