Poorly optimized thermal cycling conditions are known to produce artifacts during the polymerase chain reaction (PCR). The authors report on a problem encountered while developing a PCR protocol that may occur frequently in other laboratories, but may go unrecognized by many investigators. PCR reactions were cycled for 20, 26, 32, 38, or 44 cycles, the products of the reactions were electrophoresed, and the resulting gel was probed by means of performing a Southern blot. The results suggested that after approximately 30 cycles, most of the PCR primers are converted into PCR product, and that 3'-OH ends of the PCR product start to anneal to the genomic template DNA. These are then extended during continued cycling, resulting in large molecular weight products that are not the desired product for DNA sequencing. The authors recommend that the total number of cycles in PCR be minimized to avoid the buildup of such nonspecific products. The results presented here support the recommendation and also demonstrate that excessive cycling can lead to a dramatic loss of specific PCR product.