Many restoration philosophers, Aldo Leopold and William Jordan among them, use the metaphor of repairing a broken machine (such as a clock) to explain how restoration can serve not only to repair the damaged ecosystem, but also to provide us with a unique opportunity to understand how an ecosystem works (Jordan et al. 1987). A clock repairer, or mechanical ‘tinkerer’ of any sort, often finds out how to fix a machine by pulling it apart and putting it back together again. Without a blueprint, various experimental configurations may be required before one configuration (or perhaps more) results in successful function. Attempting to repair a damaged ecosystem, it follows, can provide us with new opportunities to study the (often not readily observable) healthy workings of natural systems; including which structural elements are critical or how the processes and interactions are ordered. Restoration can thus provide opportunities for analysis and experimentation that are either not available in healthy ecosystems or simply could not be countenanced if deliberate damage were a prerequisite (Hobbs & Hopkins 1990). This theme was prominent in the seminal text, Restoration ecology: a synthetic approach to ecological research (Jordan et al. 1987) in which reference was often made to restoration as being the ‘acid test’ of theory. Some would argue, however, that the potential of restoration being an ‘outdoor laboratory’ for learning about how nature works is being substantially under-utilised. But is it true that ecologists are failing to take up opportunities to learn from practice? Certainly there has been a degree of inertia in Australia in past decades (less so in New Zealand in the case of learning from island restoration), but it would seem to me that there is now growing interest in using restoration to test ecological concepts. Learning from restoration. In this issue of EMR (and lined up for publication in future issues) we provide a number of reports of projects that have been deliberately set up not only to repair the structure and function of sites but also to provide more insight into the workings of particular ecosystems or aspects of them. Anne Jensen’s feature on repairing wetlands in the lower Murray Valley, for example, describes a deliberate effort to trial restoration approaches derived from science, with results being monitored by research partners. This is being done not only to further refine restoration practice but also to provide further insight into wetland process and function. The second feature (reporting restoration trials in Blue Gum high Forest in northern Sydney) also examines the responses of a site to a range of treatments. The stated objective is not only to improve restoration practice but also to improve our basic understanding of the ecological property of resilience to natural disturbances. In both cases, healthy reference sites are not available and, even if they were, it is doubtful whether they would be sufficiently ‘transparent’ to act as models for repairing degraded systems. Nor is any one system the same. We often need to ‘prod and poke’ the actual sites themselves to understand the structure and dynamics of the pre-existing system. Like the clockmaker repairing an old clock who’s origin and workings is unknown, restorationists test by trial and error whether the community can recover in response to our informed guesses or whether we need to discover new triggers or missing components to render a dysfunctional system functional again. In terms of learning from science, this issue’s technical report (Brown et al. p. 28) demonstrates how experimental method is needed to adequately assess the most effective technique for restoration practice. Identifying appropriate treatments for control of individual weed species in bushland is not carried out as often as we would like, largely due to tight practitioner budgets and the partitioning of operations into research on one hand and practice on the other. Rarely can funding be gained for sound, controlled trials, although it is notable that this project was funded through an NHT grant to enable the employment of a coordinator to advise on experimental design and guide volunteers in consistent treatment and monitoring. The two research reports in this issue are also strongly linked to practice. Nest boxes are devices used for surveying fauna per se (Smith & Agnew p. xx) but they can also be used to provide habitat ‘bridges’ until a ‘restored’ ecosystem is mature enough to provide its own nest hollows. In this way, research and practice can be mutually beneficial and the differences between them seamless. The information provided in Cochrane et al. (p. xx), focuses on germination requirements for the purposes of improving threatened species restoration practice, but could be equally useful to understanding germination triggers in undamaged ecosystems; enhancing our potential to better manage them.