Abstract: Tail-biting and other harmful social behaviours are a common problem on pig farms. The aims of the current experiment were (1) to investigate the genetic component of harmful social behaviours such as tail-biting by assessing breed differences, and (2) to further investigate the reliability and predictability of a test, [`]the tail-chew test', previously identified as potentially being capable of predicting a pig's predisposition for tail- and ear-biting. The behaviour of three pig breeds (Large White (LW), Landrace (LR), Duroc (DR)), with 100 pigs per breed, was observed in a [`]tail-chew test', and by observing the performance of harmful social behaviour directed to pen mates in flat deck pens after weaning. The tail-chew test, carried out on two consecutive days pre-weaning, involved observing the behaviour of individual pigs towards two suspended ropes. Pigs were weaned at 28 days and the occurrence of harmful social behaviour was recorded 4 weeks later over 2 consecutive days (1 h per day) using a group [`]period occurrence' scanning method. Breed had a significant effect on rope-directed behaviour in the tail-chew test and on harmful social behaviour. DR pigs interacted with the ropes in the tail-chew test more often (median 23.0 vs 19.0 and 17.5 times in 20 min, P<0.001) and for longer (31.0 vs 20.0 and 23.2 s, P<0.001) than LR and LW pigs, respectively. Although not significantly different from LW, DR pigs tended to direct more total harmful social behaviour towards pen mates than the other breeds. In particular, DR were observed in more total pig-directed biting of pen mates (median 9.0 vs 6.0 and 7.0, P<0.01) than LR and LW, and tended to nose pen mates more often than the other breeds (13.0 vs 11.0 and 10.0, P=0.06). LR pigs bit the ears of pen mates less often than LW and DR (4.0 vs 5.0 and 6.0, P<0.001). Belly-nosing activity was low, with a median of 0 for all breeds, but LR belly-nosed pen mates more often than Durocs (interquartile ranges 0-2.0 vs 0-1.0 and 0-1.0, P<0.01). The behaviour observed in the tail-chew test on day 1 correlated significantly with that observed on day 2 of the test (e.g. frequency of rope-directed behaviour rs=0.380,P<0.01). There were significant but weak correlations between rope-directed behaviour and the performance of some harmful social behaviours. The significant breed differences indicate some genetic contribution to expression of harmful social behaviours. However, the tail-chew test was found to be of limited ability to predict tail- and ear-biting under commercial conditions.