Resumo. One of the main tools to quantify and to qualify the forest potentials is the forest inventory. During its planning phase it is necessary to define the sampling intensity as well as the sizes and shapes of sample units to minimize data collection costs while increasing the precision of the forest inventory. This study aimed to test different sizes and shapes of sample units for different diameters at breast height (DBH) size categories. The study was carried out in the Manaus region including INPA and EMBRAPA experimental station areas around BR-174 road, in the state of Amazonas, Brazil. Five categories of DBH were considered, as following: above 5cm, 10cm, 20cm, 25cm and 45cm. For all DBH categories, twenty-three different sizes were tested, from 100m2 to 10,000m2. The number of samples was defined as n = 30 for all size simulations to qualify for the use of the central limit theorem assumptions. Basal area was used to estimate uncertainty in relation to the true value of mean for comparison among different sizes for each DBH category. Almost all sample plot sizes in all categories of DAP obtained uncertainties below 10%. The best sample sizes were 1,000m 2 , 800m², 1,200m², 2,000m² and 10,000m², respectively for minimum DBH of 5, 10, 20, 25 and 45cm. Most of exploratory forest inventories in the state of Amazonas have been done for trees with DBH greater than 10cm, and sizes have varied from 1,000 to 2,000 m2. The findings of this work demonstrated that the size for exploratory inventory is adequate. However, different size should be used for DBH greater than 45cm, which is usually the minimum DBH for logging plans.