Monitoring the success of rhizobial inoculation requires reliable identification of the introduced strains in nodules and when recovered from field soil. The polymerase chain reaction (PCR) coupled with the use of either random or directed primers has increasingly become the molecular method of choice for characterising bacteria at the strain level. We have investigated the use of 5 markers (REP, ERIC, BOXA1R, RPO1 and IGS) commonly used for PCR fingerprinting to characterise rhizobia bacteria used in the manufacture of rhizobial inoculants in Australia. PCR with random primers often yields inconsistent results because most protocols do not specify stringent cycling and non-cycling parameters. We have increased the stringency and improved the specificity of reaction conditions for 4 of the 5 markers tested. Optimised protocols were then used to fingerprint the 39 strains of rhizobia bacteria held in the 1998 mother culture collection of the Australian Legume Inoculant Research Unit (ALIRU). Results for 34 strains using at least one marker are presented. Although the mother cultures of these inoculant strains undergo numerous quality assurance tests annually, it was not until PCR fingerprinting was applied that 2 strains, believed to be unique, were found to be identical. In the subsequent investigation, we determined that the 2 strains were originally unique but that a mix-up in the cultures had occurred at least 3 years before our analysis. Use of serology, plant infection tests and field tests were not sufficient to detect this problem. The use of PCR fingerprinting with optimised protocols has now been incorporated into the annual quality assurance regime used by the ALIRU who monitor strain quality for the Australian rhizobial inoculant industry. Higher quality rhizobial inoculant for use by Australian farmers is a beneficial outcome of this work.