The work was carried out in ten Seed Testing Laboratories, located in the State of São Paulo, Brazil. Ten commercial seed lots of guinea grass were used in order to evaluate if the seed sample dividers were or not responsible for the differences of results in the purity analysis. The analysis were developed in 6 steps. In steps A and D, the seed samples dividers available in each laboratory were used; in steps B and E, the soil divider was used; and the Boerner divider was used to obtain the work samples in steps C and F. For the development of steps A, B and C, the purity analysis were carried out in each of the 10 laboratories. However, for the development of steps D, E and F, the purity analyses were carried out in only one laboratory. In step D, the work samples were obtained in each laboratory using the available seed sample divider. After that, the purity analysis were carried out in one laboratory. The results showed that the seed sample dividers available in the laboratories, as well as the soil divider, were responsible for large variation of results between laboratories; that the Boerner divider showed the smallest variation of results; that the results of the purity analysis using a soil divider, although not significativelly different of the results obtained by the Boerner divider, showed a larger variation of results than the Boerner divider.