Introduced species of mammals have now been removed from many islands around New Zealand, thus providing singular opportunities for ecological restoration. If island restoration is to be attempted, the way island biota originate and the precise effects of introduced organisms must be identified. Plants introduced to the New Zealand archipelago may have transitory effects, but others may modify forest structure and disrupt succession. Goats have been the most destructive introduced herbivore on islands. Among introduced predators, cats have extirpated colonies of seabirds, and rats (depending on species) affect invertebrates, lizards, and birds. Ecological theories and concepts that may help with island restoration projects include: the keystone species concept, in which the effects of one species on others is disproportionate relative to its abundance; the "intermediate predator" hypothesis, where removal of the top introduced predator may lead to rebound effects of intermediate predators; and ecological chain reactions, where local extinction of some species can cause complicated multiple effects. Problems with restoration of islands may be encountered because of meagre data on the previous effects of pests (such as predators), use of non-seral species in revegetation projects, proliferations of indigenous or introduced species that have unforeseen community effects, and inexplicable difficulties with some translocations. A restoration case study in the continental Mercury Islands and on Cuvier Island showed success with removal of introduced mammals and demonstrates the various effects of introduced browsers, grazers, and predators. A contrasting case study is provided by oceanic Mangere Island in the Chatham Islands where 22 species of avifauna have been lost, seven as permanent extinctions. Restoration targets for some New Zealand islands can be clarified by palaeoecological studies of Maori (Polynesian) middens and natural deposits. Understanding the role of disturbance in island systems may also help clarify restoration targets. When exotic keystone species are introduced, physical disturbance may be overridden by biotic disturbance. This replacement in turn has implications for trophic structure. With high levels of biotic disturbance, continental islands may be changed from relatively species-rich bottom-up food webs to species-poor top-down trophic cascades. These possibilities can be tested with an experimental approach to restoration, although such experiments may be hard to interpret because of difficulties with replicates and controls. Ecological restoration on New Zealand islands has potential to replace damaged or lost communities, expand the ranges of relict populations, reduce the selective influence of exotic (keystone) species on indigenous species, help in understanding how the systems are formed, provide opportunities for educational and scientific investigation, and act as a testing ground for new technologies against pests.