Risk taking, as is any other phenotypic and/or behavioural trait, is determined by proximate constraints related to time or resource availability and by evolutionary adaptive restraints related to the differences in the costs of risk taking and its benefits in terms of fitness. Because risk taking is influenced by many confounding variables related to experimental design, environment, parents and offspring, few field studies have been reported which unambiguously separate the effects of restraints from those of constraints. We compared parental risk taking in blue tits (Parus caeruleus) during brood defence towards a nest predator in broods with experimentally reduced and natural egg-hatching success leaving the original number of eggs in the nest. The experimentally reduced broods had more time or resources available and lower risk-taking benefits compared to the control broods. 'Constraint' would predict more risk taking in broods having experimentally reduced egg-hatching success, whereas 'restraint' would predict the opposite effect with more risk taking in broods with natural egg-hatching success. We report, to our knowledge, the first field study experimentally demonstrating a brood defence restraint in response to reduced egg-hatching success. This demonstration was only possible after controlling for more than 20 potential confounding variables showing once more how complicated it is to separate proximate from evolutionary levels of analyses in natural populations.