1. Until recently, neither the phycitid moth Cactoblastis cactorum nor the cochineal insect Dactylopius opuntiae have been satisfactory biological control agents of Opuntia stricta in South Africa. 2. In marked contrast, both of these agents have kept O. stricta under biological control for many decades in Australia. 3. In an attempt to improve the situation in South Africa, a stock of D. opuntiae was obtained from O. stricta in Australia during 1996. 4. Host‐specificity tests confirmed that the newly imported D. opuntiae from Australia is a different biotype to the one already established in South Africa. 5. The Australian (‘stricta’) biotype thrives on O. stricta but is unable to develop satisfactorily on O. ficus‐indica, while the converse is true for the South African (‘ficus’) biotype, which thrives on O. ficus‐indica but fares poorly on O. stricta. 6. The integrity of the host‐plant specificity of the two biotypes of D. opuntiae has important implications for biological control of Cactaceae in South Africa, and has greatly enhanced prospects that O. stricta can be brought under biological control successfully.

Summary 1.Results of recent research on Dactylopius opuntiae , a biological control agent for cactus weeds ( Opuntia spp.) in South Africa and elsewhere, challenge the maxim that genetic diversity of agents necessarily enhances the chances of success in biological weed control. 2. Two biotypes of D. opuntiae , each specific to a different Opuntia species, interbred freely, at least under insectary conditions. We therefore carried out cross-breeding experiments to determine the viability and host-preferences of progeny produced by these crosses. 3. Unlike their parents, F 1 hybrids were not species-specific, developing equally well on either of the parental hosts, Opuntia ficus-indica and Opuntia stricta . The situation was more complex in F 2 back-crosses between hybrids and in crosses between parent strains and hybrids because male cochineal insects contributed only maternally inherited genes to their progeny, due to their unusual haploid-diploid (lecanoid) mechanism of sex determination. Some F 2 combinations produced cohorts of progeny that were either entirely true-bred (i.e. host-specific) or entirely hybrid (i.e. not host-specific) genotypes, while other combinations produced groups of siblings with some individuals (theoretically half ) that were true-bred genotypes and the balance were hybrid genotypes. 4. The lack of host-specificity of hybrids should enhance overall biological control of the target species directly, because hybrids attack both host-plants, and indirectly, because hybrid nymphs have greater chances of finding a suitable host-during passive dispersal. However, this advantage will be negated when F 2 crosses produce host-specific nymphs on host-plants that are incompatible for their survival. 5. These findings show that only pure strains of D. opuntiae should be released in monocultures of the target weeds. More generally, they caution that the possible consequences of mixing genotypes of a biological control agent species should be investigated before different provenances are amalgamated to enhance genetic diversity.