We investigated the effect of time of season on the accuracy of species number estimation from point-count data collected at 28 oases from southern Tunisia. Each oasis was visited at the beginning of the breeding season and two months later, which allowed us to conduct counts on five points per oasis per visit. For each oasis, we considered the observed species number as the total number of species recorded during each visit, and we used a capture-recapture approach to estimate species number from patterns of presence and absence of species over the five points. We found that birds were more detectable at the beginning of the breeding season than two months later, and the observed species number showed a significant decrease between the two periods. However, when a capture-recapture approach accounting for heterogeneity in species detectability was used, similar estimates were obtained from data collected during both periods. Nonetheless, the estimates obtained at the beginning of the breeding season were more precise than those obtained later. Overall, our results illustrate once more the need of taking into account time of season as an important source of bias when attempting to determine species richness from count data, and stress the need for using probabilistic approaches in such an investigation.