In an attempt to test predictions of the optimisation hypothesis of life history traits in birds, we estimated fitness consequences of brood size manipulations. Experiments were carried out over a period of 4 years in a Mediterranean population of blue tits Parus caeruleus which is confronted with a particular set of environmental constraints. Effects of brood size manipulation were investigated in relation to year-to-year variation in environmental conditions, especially caterpillar abundance. There was a strong variation in the effects of brood size manipulation depending on year. Most effects were on offspring quality (fledging mass, tarsus length). The absolute number of recruits did not significantly differ among categories (reduced, control, enlarged broods) but varied considerably among years. Females recruited from enlarged broods were of lower quality, started to breed later and laid fewer eggs than those recruited from control and reduced broods. Neither parental survival nor reproductive performances of adults in year n + 1 was affected by brood size manipulation in year n. Thus there was no evidence for a cost of reproduction in this population. Since the number of recruits did not depend on brood size manipulation (recruitment rates were higher in reduced broods), but recruits from reduced broods were of better quality compared with other groups, we conclude that adults lay a clutch that is larger than that which is predicted by the optimisation hypothesis. Producing more young could incur some penalties because offspring from large broods are of lower quality and less likely to recruit in the population. Two possible reasons why decision rules in this population seem to be suboptimal are discussed.