The way that variation in paternity affects the optimal level of paternal effort has been a contentious issue, both in terms of theory and the empirical data needed to test competing theories. Clarification of the theoretical issues has led to the prediction that a reduction in paternal effort should only be expected when (i) there are substantial costs of paternal care and (ii) males have available some cue to their share of paternity in the current brood. Previous work on the collared flycatcher, Ficedula albicollis, has shown that the first condition is supported because of trade–offs between paternal effort and secondary sexual character size. We carried out experimental manipulations of pairs of collared flycatchers (temporary male removal), which were effective in causing variation in paternity in the current brood. Male responses to these manipulations were studied by quantifying levels of paternal care. All males reared nestlings cross–fostered from non–experimental nests at the egg stage, thus ruling out the possibility that they responded to direct cues about paternity. The timing of male removal predicted the male's share of paternity, suggesting that males had a clear cue to their share of paternity, thereby fulfilling the second condition. As expected, the male's share of care, and rate of provisioning, were positively related to his share of paternity. The suggestion that the timing of removal was the cue used by males to predict their share of paternity was supported, since after the influence of this variable was controlled, there was no longer any relationship between paternity and paternal care. These data provide qualitative support for optimality models of paternal care in relation to certainty of paternity, and suggest that quantitative tests of the models are possible in well–characterized systems.