The results of the Panicum maximum Jack seed analysis show a great variation when samples from the lot are analysed in different laboratories. The centrifugal and soil dividers, used to obtain working samples, were compared, to verify the magnitude of the error introduced in the analysis results, through these equipments. Samples from two seed lots, one of clean seed and other unclean, were tested. Were also compared, the working samples obtained from the left and from right side of the divider. The centrifugal divider was operated in two ways: a) by a continuous operation; b) turned on at each time only after the seeds were poured in the hopper. On each working sample the percentage of pure seeds was determined. The analysis of variance did not show signifïcant differences between the procedures. Although not statistically signifïcant, it was observed that clean seeds in the working samples taken from the left side of the divider tended to contain higher percentage of pure seed.