T-RFLP is a newly developed molecular technique for microbial diversity analysis. Compared with other techniques, T-RFLP has several unparalleled advantages. However, there is no report on measuring the system errors generated in T-RFLP operational process. In this study, diversity analysis of nifH gene from 4 soil samples which represented different invaded stages after Ageratina adenophora (Sprengel) invasion was chosen as a case. In order to determine possible system errors in our study, multiple analysis in which only one procedure of the whole operational processes was changed each time were conducted. Results showed that repeatability for the results of T-RFLP analysis was influenced mostly by the types of Restriction Endonucleases, followed by the PCR procedure, while capillary electrophoresis has almost no effect.