We frequently use competitive PCR in the plateau phase in quantifying DNA species with a small number of cells. However, the basic issues of this method are poorly understood. Here, first we analyze this method theoretically under a generalized condition that competitor and target DNA products accumulate with different amplification efficiencies. We show a theoretical reason that competitive PCR might quantify DNA more accurately during the plateau phase than during the exponential phase. Second, we demonstrate that the theoretical predictions are supported by the experimental results of beta-globin gene amplification using the lysates of human diploid fibroblast WS1 cells. We also demonstrate that we can correctly quantify target DNA by keeping the starting concentration of target DNA close to a constant preset value while using a constant number of PCR cycles and by using WS1 cells as control. Finally, we show the experimental errors in routine measurements of c-myc copy number/cell in human leukemia HL-60 cells with various levels of c-myc multiplication. The number of c-myc copies/cell was determined with an error rate of less than 10%, where agarose gel bands were stained with ethidium bromide for the product quantitation.