Size-based community and ecosystem metrics, such as mean body mass and the slopes of size spectra, have been proposed as indicators to support the Ecosystem Approach to Fisheries Management (EAFM). These metrics show relatively consistent responses to size-selective exploitation, and "unexploited" indicator reference points may be predicted with models of size-structured foodwebs. Whereas unexploited reference points provide a baseline for assessing the relative magnitude of fishing impacts, target or limit reference points are needed to guide management. Values for target or limit reference points are difficult to justify on scientific grounds. However, given that fishing impacts in most ecosystems need to be reduced to meet the objectives of the EAFM, we argue that reference directions provide alternative medium-term management targets. We show that the power of surveys to detect trends consistent with reference directions depends on the range of body size classes included in the analysis. Selection of different size ranges will weight metrics to respond to the release of small fish from predation, the depletion of larger individuals as a consequence of exploitation, or both. Such weightings may not be consistent over time, because the differential vulnerability of larger species, within-population changes, predator-prey relationships, and the effects of competition depend on contemporary rates of fishing mortality and the history of exploitation. The power of the surveys investigated is poor on time scales of less than 5-10 years. Therefore, size-based indicators provide better support for medium-term, rather than year-on-year, management decision making. (c) 2004 International Council for the Exploration of the Sea. Published by Elsevier Ltd. All rights reserved.