Articles The

recent literature in ecosystem ecology revealstwo parallel trends. The first is an increased recognition of the importance of time and space scales as determinants of the patterns researchers observe and predict. This is evident in the steady increase in the number of journal articles listing scale as a keyword (figure 1a; Schneider 2001) and in the publication of numerous recent books devoted to scaling theory (e.g., Peterson and Parker 1998). The second trend is the increased use of manipulative field- and laboratory-based experiments to test ecological theory (figure 1b; Ives et al. 1996). Although scale and experimentation are conceptually linked, they remain awk-wardly disconnected in practice; to date, experimental work in ecology has not taken full advantage of, or been effectively used to advance, scaling theory. Consequently, a debate has developed among ecologists regarding the appropriate scales for conducting ecosystem-level research. It is commonly recognized that the scales selected for an investigation reflect a delicate balance among control (the abil-ity to relate cause and effect, to replicate, and to repeat experiments), realism (the degree to which results accurately mimic nature), and generality (the breadth of different systems to which results are applicable; Kemp et al. 1980). Enclosed experimental ecosystems (micro- and mesocosms) have gained in popularity because they maximize opportu-nities for control (Odum 1984). This is particularly true in coastal aquatic habitats, where options for controlled field manipulation are constrained by complex bidirectional exchanges between land and sea. However, a number of critics have suggested that the high degree of control afforded by mesocosms comes with a concomitant reduction in realism (e.g., Schindler 1998). Others point out that problems of realism and scale are likewise evident in whole-ecosystem manipulations (Fee and Hecky 1992); both in mesocosms and in the field, researchers must extrapolate results from smal