Resumo. It is usual to find papers about seed germination of native species using different sample sizes. However, the consequences of these procedures on the results are unknown. This study measured the germination process of Bowdichia virgilioides based on different seed sample sizes. The assay was conducted in a germination chamber under continuous fluorescent white lamps and 26.2 ± 2.5 oC. The design was completely randomized, as a 3 x 3 factorial (three sample sizes x three seed lots), with four replications per treatment. The sowing was done in gerbox, each one containing 25 seeds, making up 100, 200 and 400 seeds per sample analyzed. Germinability, time (first, average and final), speed (mean germination rate and Maguire’s index – VE or IVG), coefficient of variation of the germination time, uncertainty and synchrony of the germination process were evaluated. Those seeds with embryo protrusion were considered germinated. Germination relative frequency graphics were also made. Among these measurements, the most fragile to size sample were VE and the uncertainty. VE was also influenced by the relation between germinability and mean germination rate, demonstrating its instability to measure the speed of the germination process when the lots have different germination capacity. Results of time measurements, uniformity, mean germination rate and germinability were not affected by seed sample size. Germinability, however, was the most stable germination measurement, and was not affected by sample size, even when analyzed by seed lots with very discrepant quality.