A simulation study is carried out in order to compare the performance of the Lincoln-Petersen and Chapman estimators for single capture-recapture or dual system estimation when the sizes of the samples or record systems are not fixed by the researcher. Performance is explored through both bias and variability. Unless both record probabilities and population size are very small, the Chapman estimator performs better than the Lincoln-Petersen estimator. This is due to the lower variability of the Chapman estimator and because it is nearly unbiased for a set of population size and record probabilities wider than the set for which the Lincoln- Petersen estimator is nearly unbiased. Thus, for those kind of studies where the record probability is high for at least one record system, such as census correction studies, it should be preferred the Chapman estimator.