The successful biological control of the shrub-like prickly pear Opuntia stricta Haworth in Australia by cactoblastis cactorum (Berg) was not repeated when C. cactorum, derived from the Australian population, was released in South Africa in the 1930's against the tree prickly pear Opuntia ficus-indica (L.) Miller. Resistance of the woody portions of o. ficus-indica to attack by C. cactorum was regarded as the main reason for the poor performance of C. cactorum in South Africa. C. cactorum also oviposits and feeds on Opuntia aurantiaca Lindley, which is currently South Africa's most important weed and which is also considered to be partly resistant to attack by C. cactorum. This study had three main objectives: (i) to compare the ecology and effectiveness of C. cactorum as a biological control agent on O. ficusindica and O. aurantiaca; (ii) to reassess why C. cactorum has not been as effective a biological control agent in South Africa as it has been in Australia; and (iii) to evaluate whether inundative release or the importation of new biotypes of C. cactorum from South America (where it is indigenous) might be feasible methods of improving its effectiveness as a biological control agent of O. aurantiaca in South Africa. All field work was undertaken at a site near Grahamstown in South Africa. The ecology and effectiveness of C. cactorum on O. ficus-indica and O. aurantiaca was assessed in terms of its oviposition behaviour, survival and feeding on these host plants. The proportion of C. cactorum eggs laid on O. ficus-indica and O. aurantiaca was similar and was influenced by the size, conspicuousness and condition of the host plant as well as by the proximity of the host plant to moth emergence sites. Factors affecting oviposition site selection on the plant are also considered. Life tables, compiled for a summer and a winter generation, showed that the survival of C. cactorum was greater on O. ficus-indica than on O. aurantiaca, mainly because higher egg predation by ants occurred on the latter host plant species. During the period of study, the population size of C. cactorum was reduced by a number of mortality factors, of which egg predation and the effects of low temperatures on fecundity were the most important. Although there was evidence of a partial, positive response by predatory ants to C. cactorum egg densities on plants, the extent of egg predation was also affected by other factors, particularly seasonal effects. C. cactorum destroyed a greater percentage of cladodes on O. ficu-indica than on O. aurantiaca, but even on O. ficus-indica it was unable to contain the growth of plants within the study area. C. cactorum larvae rarely killed the woody rooted cladodes of O. ficus-indica and O. aurantiaca and consequently whole plants were not often destroyed. The detrimental effects of host plant resistance, natural enemies and climate on the effectiveness of C. cactorum as a biological control agent all appear to be greater in South Africa than in most of the regions occupied by C. cactorum in Australia. A field experiment conducted at the study site showed that inundative release methods for improving the effectiveness of C. cactorum on O. aurantiaca are not feasible. The importation of biotypes of C. cactorum from South America that might be better suited for destroying O. aurantiaca infestations in South Africa, is also not a viable option. Results of a survey of a 218 ha area that is regarded as being heavily infested with O. aurantiaca, illustrate how this cactus species has been overrated as a weed problem. It is argued that the present strategy for O. aurantiaca control in South Africa is not based on sound economic or ecological criteria.