Baffles contained in conventional actuators may be a convenient alternative to some of the extension devices used presently with metered-dose inhalers (MDIs). Actuators were modified to determine whether baffles could be used to decrease the output of large "nonrespirable" droplets. These actuators were tested using a series of nonaqueous suspension aerosols containing 0.1 to 2.0% micronized disodium fluorescein (DF) as the model drug, stabilized by sorbitan trioleate in a constant blend of fluorocarbons 11, 12, and 114. A 25-microliter metering volume was used throughout. Aerosol output was characterized by cascade impaction. Baffle size and position had pronounced effects on actuator retention and aerosol output. Increasing baffle size resulted in increased retention in the actuator. The total output of the MDI in the "respirable" range (aerodynamic diameter, Dae, less than 5.5 microns) was greater in the unbaffled actuator than in all baffled actuators. However, all baffles increased the respirable fraction (DF with Dae less than 5.5 microns: total DF leaving the actuator), R, when compared to their unbaffled controls. For example, for a 0.1% DF, 0.14% surfactant formulation, R was increased from 0.40 (unbaffled) to 0.71 by incorporation of a 0.6-cm-diameter sphere 1.3 cm from the jet of the actuator. In these cases, aerosol segregation occurred due to droplet inertia in the high velocity gas flows. Increasing the respirable fraction at the expense of the total respirable output may obviate undesirable clinical effects.