The simplest model for pump-mediated multidrug resistance is elaborated quantitatively. The way in which toxicity data should be evaluated to characterize most effectively the drug-efflux pump is then examined. The isotoxic drug dose (D10) depends on too many unrelated properties. The D10 of a cell line taken relative to that of the parental (nonresistant) cell line has been called the relative resistance (RR). This is inappropriate for characterizing the drug pump, as it depends on the extent of amplification of the latter. The reduced RR (RRR) is newly defined as the ratio of the (RR - 1) for one drug to the (RR - 1) for a different drug. This RRR should be independent of both the drug-target affinity and the extent of amplification of the drug pump in cell lines belonging to a family. The RRR depends on the avidities with which the pump extrudes the drugs relative to the passive membrane permeabilities of the latter. In plots of RRR for one drug combination vs. that for a second drug combination, cell lines that have the same pump amplified should cluster, whereas those with amplification of (functionally) different drug-efflux pumps should segregate. Both a set of new experimental data and literature results are discussed in terms of RRR. RRRs discriminate between human MDR1 and mouse mdr1a and mdr1b, between hamster pgp1 and a mutant thereof, as well as between human MDR1 and a mutant thereof. RRRs are not affected by changes in membrane surface area. Our results indicate that RRR may be used to (a) characterize drug-resistance mechanisms and (b) determine which drug-resistance mechanism is operative. Moreover, our analysis suggests that some of the reported phenotypic diversity among multidrug-resistant cell lines may not be due to diversity in the resistance mechanism.