Summary1. Natural enemies may reduce the effectiveness of weed biocontrol agents and can also cause environmental damage, for example to a shared native insect host through apparent competition. Indeed, successful biocontrol may rely on enemy-free space and avoidance of apparent competition in the area where the biocontrol agent is introduced. 2. We surveyed parasitism in 28 insects released for weed biocontrol in New Zealand (NZ). We reviewed the global literature and databases to complement this survey, and to collate records of these insects being parasitized in their area of origin. We also collated records of native insects that feed on weeds targeted for biocontrol in NZ to test Lawton's (1985) hypothesis that, to find enemyfree space, selected agents should 'feed in a way that is different' and 'be taxonomically distinct' from native herbivores in the introduced range. 3. We found that 19, mostly native, parasitoid species attack 10 weed biocontrol agents in NZ, of which 15 were confined to five agents that possessed 'ecological analogues', defined as a native NZ insect that belongs to the same superfamily as the agent and occupies a similar niche on the target weed. Parasitoid species richness in NZ was positively correlated to richness in the area of origin. However, only agents with ecological analogues contributed significantly to this pattern. 4. A review of NZ weed biocontrol programmes indicated that parasitism is significantly associated with the failure of agents to suppress weed populations. 5. Synthesis and applications. Although our conclusions are based on an unavoidably limited data set, we conclude that biocontrol agents that escape attack from parasitoids are more likely to suppress weed populations and should be less likely to have significant indirect non-target effects in food webs. Biocontrol practitioners can reduce the chance of weed biocontrol agents attracting species-rich parasitoid faunas after introduction by (i) selecting agents that have species-poor parasitoid faunas in their area of origin, and ⁄ or (ii) avoiding agents that have 'ecological analogues' awaiting them in the introduced range.

Gorse seed weevil,Apion ulicis Forst, was released into New Zealand in 1931 to help control gorse, Ulex europaeus L. This study examines the effect of the weevil in reducing annual seed production of gorse at three sites over 3 years. Weevils oviposited only during spring, and infested up to 90% of immature pods during that period. However, the many pods produced both earlier and later than this escaped attack. Lack of synchrony between weevil reproductive activity and gorse pod production limits the impact of the weevil on the annual gorse seed crop. Better adapted populations of Apion ulicis may exist in southern Europe and these could be introduced to increase predation on seed.

Gorse (Ulex europaeus L.) has become naturalized in at least 50 countries outside its native range, from the high elevation tropics to the subantarctic islands and Scandinavia. Its habit, adaptability and ability to colonize disturbed ground makes it one of the world's most invasive temperate weeds. It is 80 years since New Zealand first initiated research into biological control for gorse. This paper briefly reviews the progress made worldwide since then, and examines future opportunities for biological control of this weed. The range of available agents is now known, and this list is critically assessed. Ten organisms have been released variously in six countries and islands and their performance is reviewed. In most cases, agent populations have been regulated either from 'top-down' or 'bottomup', and there is no evidence anywhere of consistent outbreaks that could cause significant reduction in existing gorse populations in the medium term. Habitat disturbance and seedling competition are important drivers of gorse population dynamics. Existing agents may yet have long-term impact through sublethal effects on maximum plant age, another key factor in gorse population dynamics. Along with habitat manipulation, seed-feeding insects may yet play a long-term role in reducing seed banks below critical levels for replacement in some populations. In the short term, progress will rely on rational and integrated weed management practices, exploiting biological control where possible.

Summary 1.A seed-feeding biocontrol agent Bruchidius villosus was released in New Zealand (NZ) to control the invasive European shrub, broom Cytisus scoparius, in 1988 but it was subsequently considered unable to destroy sufficient seed to suppress broom populations. We hypothesized that an invasive mite Varroa destructor, which has caused honeybee decline in NZ, may cause pollinator limitation, so that the additional impact of B. villosus might now reach thresholds for population suppression. 2. We performed manipulative pollination treatments and broad-scale surveys of pollination, seed rain and seed destruction by B. villosus to investigate how pollinator limitation and biocontrol interact throughout the NZ range of broom. 3. The effect of reduced pollination in combination with seed-destruction was explored using a population model parameterized for NZ populations. villosus is expected to eventually reach 73% (the average rate observed at the localities adjacent to early release sites). 5. The model demonstrated that 73% seed destruction, combined with an absence of honeybee pollination, could cause broom extinction at many sites and, where broom persists, reduce the intensity of treatment required to control broom by conventional means. 6. Nevertheless, seed rain was predicted to be sufficient to maintain broom invasions over many sites in NZ, even in the presence of the varroa mite and B. villosus, largely due to the continued presence of commercial beehives that are treated for varroa mite infestation. 7. Synthesis and applications. Reduced pollination through absence of honeybees can reduce broom seed set to levels at which biocontrol can be more effective. To capitalize on the impact of the varroa mite on feral honeybees, improved management of commercial beehives (for example, withdrawal of licences for beekeepers to locate hives on Department of Conservation land) could be used as part of a successful integrated broom management programme at many sites in NZ.

The safety record of weed biocontrol was questioned recently when examples of damage to nontarget plants were reported overseas Until now systematic investigations of nontarget feeding have not been performed in New Zealand Results of surveys looking for evidence of nontarget damage caused by 20 biological control agents released against weeds in New Zealand are presented Most agents (16) are apparently hostspecific However two species (Tyria jacobaeae and Phytomyza vitalbae) were recorded attacking native plants although their attack was very minor and predictable from hostrange testing performed prior to release For two other species Bruchidius villosus and Cydia succedana nontarget attack was not predicted from hostrange testing Larval feeding by these species was confined to mainly weedy exotic plants that are closely related to their target plants The reliability of hostspecificity testing and overall safety record of weed biological control in New Zealand are discussed