We assessed the pecking behaviour of caged White Leghorn hens towards feather-shaped stimuli varying in colour (red or blue), material (paper or feather) and movement (stationary or movable) attached to a board placed in the feed trough. Each of the eight stimulus combinations was presented to two replicate groups of 5 young hens for 15 min at 45 and 57 days of age. We predicted that the birds would be especially attracted to red movable feathers simulating a live bird with bloodstained feathers, Severe (forceful) pecks were directed more frequently at feather than paper stimuli (P < 0.05) and at movable than stationary stimuli (P < 0.01) but there was no differential response to red and blue stimuli. We reassessed responses to the stimuli by a subset of the original birds, now in 16 groups of four hens, at 696 and 710 days of age. We found no significant effects of colour, material or movement on the latency to peck the stimuli, or the frequency of gentle and severe pecks at the stimuli, indicating that responses to the stimulus characteristics were not consistent between young and old hens. There was a positive correlation between the frequency of severe feather pecking at flock mates and the frequency of cannibalistic behaviour (P < 0.01), consistent with reports that bleeding resulting from feather pecking can lead to cannibalism. We found no significant correlation between the frequency of pecking at the inanimate stimuli and the frequencies of pecking at the flesh and feathers of flock mates. This analysis does not take into account possible behavioural differences between primary cannibals that drew blood and secondary cannibals that joined a cannibalistic attack once blood had been drawn. We conclude that the frequency of pecking at inanimate stimuli was not a good predictor of future cannibalistic behaviour by the hens in this study. However, a tendency for future cannibals and severe feather peckers to have longer latencies to peck the inanimate stimuli warrants further investigation. It will be possible to use responses to specific types of inanimate stimuli to predict cannibalistic tendencies only if future cannibals are found to have stable responses to those stimuli over time. (C) 2000 Elsevier Science B.V. All rights reserved.