We compared direct enumeration, the closed population models of program CAPTURE, the Jackknife estimator, Chao's (1987) moment estimator, and the Jolly-Seber open population model in estimating abundance of Virginia opossums (Didelphis virginiana) and raccoons (Procyon lotor). For 57 months, we used mark-recapture and radio-telemetry methods to study populations in a 600-ha watershed in northern ... [Show full abstract] Virginia. Three hundred and sixty-one opossums and 407 raccoons were captured 2,187 times, and 156 individuals were radiocollared. We used the radio-tracking data to examine departures from the assumptions of the various estimators and to obtain an independent estimate of population size as a means for comparing the other estimators. This RADIO estimate is based on the probability of capturing animals known to be in the study area. Direct enumeration typically underestimated population size relative to the RADIO estimates. Movements by animals to outside the study area resulted in direct enumeration occasionally overestimating population size. Low probability of capture in a trapping period resulted in a pronounced negative bias of the estimates produced by the selected model of program CAPTURE. Jackknife estimates were more correlated with the RADIO estimates than those from CAPTURE, but daily capture probabilities were extremely low. Chao's moment estimates were poorly correlated with RADIO and had the highest coefficient of variation of all estimators. Jolly-Seber estimates were highly correlated with RADIO and exhibited the least deviation from RADIO for both species. Goodness-of-fit tests of the Jolly-Seber model indicated that the model's assumptions were violated in some months. The cause and direction of the bias could be ascertained for these months. The standard errors of RADIO, CAPTURE, Jackknife, Chao, and Jolly-Seber estimates were all highly correlated with estimate size.