Abstract. Tree swallows defend their nests against predators with a combination of active and passive defence behaviour. Active defence consists of repeated dives towards the predator accompanied by alarm calling, while passive defence is limited to circling above the predator, with alarm calling at lower rates than during active defence. During 224 presentations of a ferret or rat snake near nests of tree swallows, up to 13 neighbouring tree swallows joined in the defence bouts. With only one exception, these defending neighbours limited their participation to passive defence. In 66 of the predator trials, no neighbouring birds engaged in defence. These trials with no defending neighbours were at nests with significantly lower intensities of parental defence, and there were no other detectable differences between the nests with and without defending neighbours. Of the 158 trials during which defending neighbours were present, the number of neighbours recruited to a defence episode was strongly correlated with male calling behaviour and male and female attendance, with the strongest effect due to the maximal alarm call rate of the defending male parent. Within the limited range of densities provided in the experiments, the numbers of defending neighbours recruited was not significantly affected by the local density of active swallow nests, even though most defending neighbours came from nests within 75 m of the nests being defended. Neighbouring swallows engaged in passive defence at all stages in their nesting cycles, although significantly more had older chicks in their own nests when they defended. An examination of the various costs and benefits of anti-predator behaviour by neighbours suggests that the principal cost is the time taken from other activities, especially foraging for the neighbour's own offspring. The principal benefit appears to be defence of the neighbours' offspring and selves through distraction and 'moving on' of predators, respectively. Thus, self-interest alone appears to explain participation in group defence, although the lower benefits obtained by neighbours probably explains their lack of active defence. The responsiveness of neighbours to variation in the intensity of parental defence suggests that they are interpreting the behaviour of the parents at the nest being threatened as an honest signal of the magnitude of the threat, although it is not clear how such honesty might be maintained. In any event, the participation of neighbours in defence bouts amplifies the sensory impact of the anti-predator display and probably enhances its anti-predator function.