Research was conducted to investigate responses of Acacia karroo to severe defoliation, with or without debarking, in a semi-arid, subtropical savanna. A herd of 250 goats reduced available foliage by 87% in a 10.5-ha area of Acacia woodland over 3 weeks during the early dormant season (May). The goats debarked 37% of the trees. On average, bark was stripped less than halfway round a stem and 11% of the bark area was affected. Trees 1.6-2.5 m tall and 11-30 cm in basal circumference were most susceptible to debarking. A year later, there were no detectable effects of defoliation on the number of browseable shoots, internode length, shoot length, spine length, number of seed pods, neutral detergent fibre (NDF) or nitrogen (N). Defoliation tended to increase condensed tannin concentration (CT), but the effect was not significant (p > 0.05). The absence of goats during the growth season following defoliation allowed the trees to recover. Debarking had greater impacts on the trees than defoliation had. Debarking reduced the number of browseable shoots, while ringing at the base reduced CT and increased N. Debarking tended to reduce shoot and spine length, and halve the number of pods, but the effects were not significant (p > 0.05). Responses of basally ringed trees indicated rapid vertical growth when the trees were highly vulnerable to browsers and fires, suggesting strong selection for herbivore/fire tolerance. Among all treatments, excluding basal ringing, there were significant variations in NDF and CT between years. There was a positive relationship between N and NDF, and a negative relationship between N and CT, in 1 year, but not in the other. Among basally ringed trees, CT was negatively related to N, but NDF was not. The strong trade-off between N and CT in basally ringed trees probably indicates that large changes in root:shoot ratio are required to effect definite changes in defence.