Resumo. This study had the objective of establishing limits for seed germination, purity, 1000-seed weight and humidity, aiming at the quality control of laboratory seed analysis for forest species. To that end, results of forest species seed analysis from 1997 until 2007 were obtained from the archives of the Laboratory of Seed Analysis of the Agricultural Research Foundation of the State of Rio Grande do Sul. The limits of reliability were constructed from an analysis of data normality and the use of evaluation criteria of Pseudosigma. These limits were effective in the establishment of a minimum acceptable value in order for a plot of seeds to be commercialized, proposing that the germination values inside the low confidence limits must be discarded, because they don´t present acceptable germination rates.