BACKGROUND: Studies of birds have a disproportionate representation in the literature on life-history evolution, because of the (apparent) ease with which the costs and benefits can be quantified and manipulated. During reproduction, birds frequently show a highly conserved pattern of mass change and changes in mass loss during breeding have been widely considered to be a valid short-term measure of the costs of reproduction. Experimental manipulations of the breeding attempts of birds usually argue that the presence of a response shows that a cost of reproduction exists, but there is little consensus as to how the size of these costs can be measured. RESULTS: We model this mass loss by considering how a parent can maximise its lifetime reproductive success, using a theoretical framework that is particularly suited to modelling parental care in altricial birds. If lifetime reproductive success is taken to be the sum of a parent's current and future reproductive success, we show that the exact forms of these components will influence the optimal amount of mass a parent should lose. In particular, we demonstrate that the shape of the relationship between parental investment and chick survival will lead to differing degrees of investment between parents of different initial qualities: parents with initially high levels of energy reserves could conceivably invested a lesser, similar or greater amount of resources than parents with initially low reserves, and these initially 'heavy' parents could potentially end up being lighter than the initially 'lighter' individuals. CONCLUSION: We argue that it is difficult to make predictions about the dependence of a parent's final mass on its initial mass, and therefore mass loss should only be used as a short-term measure of the costs of reproduction with caution. The model demonstrates that we require a better understanding of the relationship between mass loss and both current and future reproductive success of the parent, before predictions about mass loss can be made and tested. We discuss steps that could be taken to increase the accuracy of our predictions.