Small mammals are important elements of tropical forests but there is not a protocol for sampling these animals. In this study, we aimed at evaluating which sampling design maximizes the richness and/or the abundance of small mammals in a given area. We used data available in studies carried out in the Brazilian Atlantic forest. The variables analyzed for each study were number of marsupial species, number of rodent species, number of marsupial individuals and number of rodent individuals (dependent variables), sampling effort (trap-nights), design (grid or transect), number of strata sampled, number of nights and number of trap types used (independent variables). We did an analysis of covariance using the type of forest (evergreen or semideciduous) as the co-factor and factoring out the sampling effort to verify if the patterns of richness and abundance of species changed between these types of forests. The same analysis was done using the design as the cofactor in different forest types. Therefore, we performed analyses of variance in each forest type using the number of strata sampled, number of traps types and number of nights as factors to verify the effects of these factors on richness and abundance of the species. The capture effort was the most important variable to explain the richness and abundance of small mammals. The forest type influenced the abundance of species. Marsupials seemed to be more abundant in the semideciduous forest and rodents in the evergreen forest. RJ, Brazil.