A previously published model for respirator performance was validated by comparison with actual exposure concentrations of a respirator wearer, and the effects of particle size, fit, and work rate on the accuracy of the model were evaluated. Measurements were made using mannequins or human subjects. The test aerosol consisted of a solution of oleic-acid and sodium-fluorescein dye. The two particle sizes used were of mass median aerodynamic diameters of 0.65 and 1.42 micrometer (microm). In the mannequin procedure, a dual cartridge half mask was glued to the face with five sealable 3.7 millimeter (mm) gaps simulating facial seal leaks. All inhaled air passed through a 90mm sample cartridge. The logs of the predicted and measured percent penetrations at steady flow rates at 2, 5, 10, 20, 50, and 100 liters/minute correlated well. No statistically significant difference was seen between predicted and measured results. No significant difference was seen for four leak conditions with fit factors from 43 to over 5,000, using two breathing cycles. Eight tests completed on 16 human subjects included two fit conditions, two work rates, and two particle size distributions. A correlation coefficient of 0.76 was determined between measured and predicted values, indicating that 58% of the observed variation was attributed to the model. The remaining variation in penetration was attributed to the random nature of respirator performance and its measurement. Further statistical analysis indicated that the accuracy of the model was not affected by the value for fit factor, particle size, or work rate. Penetration for poorly fitting respirators was underpredicted. The authors conclude that the model accounts for the effects of fit, work rate, and particle size distribution.