A model describing the lognormal total distribution of contaminant levels inside a respirator (CI) is applied to assessing the probability of toxicant overexposure among a population of respirator wearers; the model accounts for between-wearer and within-wearer variability in ambient exposure levels (CO) and decimal fraction respirator penetration (P) values. The three exceedance probabilities are defined as PrI, the proportion of all CI levels over the permissible exposure limit (PEL); PrII, the proportion of wearers with an arithmetic mean CI level over the PEL; and PrIII, the proportion of wearers with a 95th percentile CI value over the PEL. PrII is considered that fraction of the population overexposed to a chronic toxicant; PrIII is considered that fraction overexposed to an acute toxicant. The behavior of PrII and PrIII over a range of exposure parameters is explored. An important observation is that a respirator-wearing population can have a substantial fraction of toxicant overexposure even though two conditions are met: (1) the P values for the population satisfy the criterion for the assigned protection factor (APF); and (2) the population arithmetic mean CO level is at or below the maximum use concentration (MUC), defined as APF x PEL. The authors recommend that the current MUCs for air-purifying respirators be reduced by one-half to reduce the potential respirator-wearing population fraction of overexposure and that appropriate exposure surveillance programs for all wearers be mandated.