Neotropical species richness is often estimated using accumulation or rarefaction curves and extrapolations based on various methods, but the efficacy of these methods is rarely tested. We used both a time-limited trapping session and previous knowledge of the mammal fauna from a site in French Guiana to validate these different methods. Three currently used extrapolation models were tested on our data. In a 2.5-mo trapping period we caught 75 individuals representing 18 small non-volant mammal species in three different habitats. We found that for comparable trapping efforts, village edges presented a higher abundance and species richness than primary and secondary forests. Species richness extrapolations using current models demonstrated that the exponential dependence model fits the known diversity of the site better, although this model is usually applied to large geographic areas and to relatively poorly documented taxa. Both Clench equation and linear dependence models underestimate small-mammal species richness in our study. We finally emphasize the interest of sampling in such edges surrounded by either primary or secondary forest, to maximize the probability of trapping rare species.