Bird song is a sexually selected male trait where females select males on the basis of song quality. It has recently been suggested that the quality of the adult male song may be determined by nutritional stress during early development. Here, we test the 'nutritional–stress hypothesis' using the complex song of the European starling. Fledgling starlings were kept under experimental treatment (unpredictable short–term food deprivations) or control conditions (ad libitum food supply), for three months immediately after independence. We measured their physiological and immune responses during the treatment and recorded song production during the following spring. Birds in the experimental group showed increased mass during the treatment and also a significantly suppressed humoral response compared with birds in the control group. There was no difference between the groups in the cell–mediated response. Next spring, males in the experimental group spent less time singing, sang fewer song bouts, took longer to start singing and also sang significantly shorter song bouts. These data support the hypothesis that both the quality and quantity of song produced by individual birds reflect past developmental stress. The results also suggest the 'nutritional–stress hypothesis' is best considered as a more general 'developmental–stress hypothesis'.