ecosystems. Thoughts from the final session, and other issues related to considerations for ecological research on GEOs raised earlier in the meeting, are summarized below. These do not represent the consensus of the group, but reflect the diversity of issues that arose during the workshop.
Scale Scale is an issue in looking at the environmental effects of GEOs, and for that reason, experiments and experimental protocols at a larger scale, from mesocosms up to landscapes, are likely to be needed, depending on the taxa and containment constraints. Many participants identified large-scale, organized collaborative projects that support different research objectives, including those on GEOs, as perhaps the only way to design and fund the scale of analysis needed. Studies on a large ecological scale can take advantage of emerging remote sensing technologies that improve traditional methods of observation.
Context The context in which GEOs are used or introduced has important implications for evaluating the relative impact of GEOs on the environment. The size and magnitude of the effects of GEOs may be determined by the system in which they are used, and that system (for example, row-crop agriculture) may itself have much larger impacts on the environment than the isolated effects of GEOs.
Comparisons, Contrasts, Baselines In evaluating the effects of GEOs on natural habitats and wildflife, selecting the appropriate comparator is critically important to the study design. It will be easier to detect subtle ecological effects if the contrast between comparative systems is sharp. The establishment of baseline states (e.g. before introduction of GEOs) can help to create the contrast needed to make appropriate evaluations of effects.
Sensitive Indicators Identifying indicator species or processes that are sensitive to specific environmental changes could assist in the detection of the effects of GEOs, especially secondary, indirect effects.
Models Models are a useful tool for studying processes that cannot be directly observed and the development of models is a research objective in and of itself. Models have limitations, of course, and need to be modified as experimental data becomes available. Nevertheless, models can identify where the most critical data needs exist, where effects are most likely to be observed, and the appropriate end points for studies looking for meaningful effects. Models can be used to envision scenarios such as the effects of multiple introductions on the ability of a species to estab-