The success in energy saving in roosting Great Tits (Parus major) is potentially influenced by roost microclimatic characteristics. We tried to learn about the pattern in cavity microclimate influence on night roosting of birds, under natural weather conditions. An experimental aviary offered two artificial tree-roosts with different microclimatic conditions which varied according to the outdoor conditions (mainly due to solar radiation) and microclimate retention capability of roosts. Birds actively explored roosts before the final choice of sleeping site. Two hours prior to sunset, there were significant differences between the roosts (n = 11 adult males tested, each six nights). Selected roosts had higher average temperature (Wilcoxon matched pairs test, P = 0.017) and temperature at the time of sunset (P = 0.028). Simultaneously, inside temperature decreased slowly (P = 0.047). During the night, when birds slept in one of the roosts, the average temperature was higher (P \ 0.001) and the fluctuation range of temperatures was smaller (P = 0.003) there. Moreover, at the morning twilight, the temperature of the occupied roost was higher (P \ 0.001). Humidity did not have a significant influence at all. The effect of metabolic heating was demonstrated. When birds slept in the insulated roost (n = 30 nights), the indoor average temperature significantly differed from the non-insulated and outdoor temperatures (Kruskal-Wallis ANOVA, P = 0.0001). Selection of an actually warmer and thermally more stable roost according to the course of weather indicates demand for a wide spectrum of potential roosts in the natural environment.