Data were analysed from 24 experiments, done between 1984 and 1988 in the Namoi Valley of NSW, Australia, in which cotton crops were subjected to various amounts of damage. There were three sets of experiments: pest management on a research station, pest management on commercial farms, and manual disbudding. The okra-leafed variety Siokra was infested with fewer pests than the normal leaf ‘DP90’ and suffered less damage from a given density of pests. Both varieties compensated for early-season damage at low yield levels but not at high levels. The reduction in yield caused by damage at high yield levels was less in ‘Siokra’ than in ‘DP90’. Three physiological explanations for these effects are offered: (1) where low yield is caused by heavy physiological shedding of fruit resulting from adverse weather or agronomic practices, the shedding caused by pests can substitute for physiological shedding and thus insulate low-yielding crops from pest damage; (2) if the yield of the low-yielding crops is limited by the incomplete interception of radiation, then delaying the termination of the expansion of the canopy will increase yield; early damage can delay the onset of the metabolic stress that terminates expansion of the crop canopy by reducing the rate at which the fruit load increases; (3) loss of early fruit under some conditions conducive to high yield can cause rank growth, whose increased respiratory requirements may reduce the assimilate supply available to set and grow the extra bolls that could compensate for those lost. The implications of the results for pest management are discussed. Thresholds should not be raised above the lower end of their current range in Australia except in ‘Siokra’ crops, which are expected to yield <7.5 bales/hectare.