The striking vocalizations, stretching, gaping and pushing that comprise the `begging' behaviour of nestling birds have been of great interest to evolutionary biologists. Many models have been proposed to explain the evolution of this suite of traits. In this paper I describe the first experimental test of the assumption that forms the foundation of all these models: that begging behaviour bears a direct cost. Miniature two-way radios hidden inside artificial nests were used to measure changes in predation rates with changes in begging. This showed that begging was costly for birds that nest on the ground, but there was no cost for those that nest in tress. Further, the cost of begging for ground-nesters was an increasing function of begging rate. For ground nests, these data suggest that nest predation could act both as a check on the evolutionary exaggeration of begging traits, and as a mechanism to enforce the honesty of begging signals. The marked variation in the predation cost of begging in different nesting sites suggests that this cost may form a good framework for investigations of the interspecific diversity of begging calls.