The results obtained by two estimators of population sizes, MNKA and Mh, were compared for four species of small mammmals - Didelphis aurita Wied, 1826, Philander frenata (Olfers, 1818), Nectomys squamipes (Brants, 1827) and Akodon cursor (Winge, 1887) - during a long-term population study. The MNKA estimator consistently underestimated the population sizes in relation to Mh. On the other, the probabilistic estimator Mh, which reduces bias through the jackknife technique, could not be used in all cases as its assumptions were not always met. Correction factors between the estimates obtained by the two methods were calculated for the last three species, for which catchability did not vary significantly in time and that presented positive correlation between the estimates by the two models. In order to combine the adavantages of both methods for small mammal population studies, is suggested the use of probabilistic closed population models and to calculate a correction factor based in another model which allow estimates in all cases, and which provides correlated estimates. This correction factors should be used in those cases where the probabilistic model cannot be used.