Plucking analysis is a standard method for assessing the breeding season diet of goshawks Accipiter gentilis. Results may be biased, however, due to differences in the detectability of remains of certain prey species and to selective prey transportation by hawks. This study investigated prey choice of three goshawk breeding pairs in the city of Hamburg, Germany, during 1997–99. Predation data obtained by continuous radio-monitoring of the adult males were used to quantify biases of simultaneous plucking analyses, which were conducted in the nesting territories of the corresponding breeding pairs. A comparison of estimated average food requirements with actual biomass intake observed in two of the monitored goshawks (with sufficient sample sizes) indicated that it was possible to record virtually all kills that occurred during tracking sessions. Therefore, radio-tracking data could be used to evaluate the accuracy of plucking analysis. Comparisons showed that the proportions of small birds and of feral pigeons Columba livia f. domestica were underestimated by the scanning method by 10% and 13%, respectively. The latter finding was unexpected as pigeons have large, conspicuously coloured feathers that are easy to detect. However, as goshawks plucked large prey more thoroughly at the capture site than smaller prey, a comparatively small number of their feathers were present in the nesting areas. This source of bias in plucking analysis has passed unnoticed in previous studies. It may have important implications for the interpretation of results of future projects that aim to estimate predation rates in goshawks and other bird-eating raptor species.