Distance sampling is usually preferred over uncorrected point counts for surveys of forest birds, but rarely has its accuracy been assessed against known numbers, particularly in tropical forests. We compared density estimates of eight species of breeding bird—Black-naped Monarch (Hypothymis azurea), Hill Blue Flycatcher (Cyornis banyumas), White-rumped Shama (Copsychus malabaricus), Puff-throated Bulbul (Alophoixus pallidus), Abbott's Babbler (Malacocincla abbotti), Puff-throated Babbler (Pellorneum ruficeps), White-browed Scimitar Babbler (Pomatorhinus schisticeps), and White-bellied Yuhina (Yuhina zantholeuca)—obtained through color-banding, nest finding, and territory mapping with those derived from distance methods in evergreen forest in northeastern Thailand. We also assessed the availability of birds to be detected using a closed-capture model and incorporated this with point-transect distance sampling. Abundance estimates from territory mapping and distance sampling were highly correlated, but biased for two species using line transects and five species using point transects. Six of the seven biased estimates were biased low. Probabilities of detection were not significantly different between lines and points, and there was no significant difference in the overall accuracy between methods. Accounting for observer differences improved density estimates but reduced precision. The variance in accuracy was mostly related to the behavior of the different species. Adjusting for availability did not improve the overall accuracy of the estimates, because of the low singing rates of tropical birds. Nonetheless, distance sampling provided relatively robust estimates despite the near total dependence on aural cues. Violations of distance-sampling assumptions may be frequent in heavily forested habitats, where both availability for detection and probability of detection on the transect line (or point) are likely to be <1.