The amount of drug that is delivered by nebulization is a combination of the physical properties of the agent being nebulized, the performance of the nebulizer, and the pattern of breathing of the patient. To avoid biological variation, mechanical models of breathing are frequently employed during the evaluation of the performance of a device. For simplicity, many investigators use sinusoidal models of breathing to calculate the expected inhaled mass, although some use square waves and other more complex models. Most assume that the duration of inspiration (Ti) is half of the total respiratory time (Ttot). This study compared the calculated inhaled mass from which the expected pulmonary deposition was estimated from the actual pattern of breathing of 43 children with cystic fibrosis (CF) breathing from an unvented nebulizer with a low dead volume and appropriate particle size distribution with that from a sinusoidal pattern of breathing using the same tidal volume (VT) and respiratory rate. The respiratory duty cycle (Ti/Ttot) was 0.45 +/- 0.05, which meant that less time was spent during inspiration than that found in a pure sinusoidal pattern. The difference between the predicted deposition from the actual pattern of breathing and that calculated from the sinusoidal model was 12 +/- 7%, which correlated with the respiratory rate (r = 0.67, P < 0.001). The degree of lung disease did not influence the discrepancy between the two values. In general, the actual VTs and respiratory rates were less in the patients than those employed in mechanical models of pediatric breathing. Although some patients had respiratory patterns that could be represented accurately with a sinusoidal model, most did not, and there were wide variations from child to child. These results suggest that there are both systematic and random errors arising from the use of a sinusoidal waveform to mimic respiratory events in patients.