may have had a prior history of cohabitation in the tropical wetlands, which may have reduced the levels of interspecific competition necessary to establish a stable system of biotic interactions within swamps. The ultimate dominance of tree ferns, with subdominant pteridosperms, is a pattern shared by Stephanian coal swamps and the clastic wetlands. We do not know if the same species occupied both kinds of habitats, but it seems unlikely, because of the edaphic constraints imposed by peat substrates. If these plants were capable of tolerating disturbances, they may have established a very different set of equilibrium interactions than those of Westphalian swamps.
In a system with organization like that of Pennsylvanian coal swamps, ecosystem persistence is a real property, empirically measurable. The biotic interactions within a system tend to conserve its ecomorphic structure and dynamics, even if there is species turnover. This generality may be true only during times when physical or extrinsic conditions remain relatively uniform for millions of years. It may also require sufficient time for a system of interactions among component species to evolve.
The concept of persistence may need to be evaluated at hierarchical levels above species composition. Although species associations can persist for millions of years, as our data suggest, these patterns may not reveal the major structural attributes of the ecosystem. To view all ecosystems as organized strictly as happenstance associations of species with similar resource requirements may reflect a bias that has grown out of studying short time intervals available to neoecology, complemented by the somewhat longer patterns of Recent palynology. The climatic fluctuations of the last 10,000 to 100,000 yr are not typical of all of Earth history (moreover, the Northern Hemisphere is not a good proxy for the tropics). The dogma of "individualism" should be reevaluated for generality.
Species with opportunistic life histories may have a significant advantage during intervals of ecosystem disruption. During a postextinction lottery, high reproductive output, high dispersibility, tolerance of a diversity of physical conditions, ability to grow in disturbed areas, and a tendency to spawn peripheral isolates (due to dispersal capacities), all favor invasive opportunists. During the Late Carboniferous this is seen in the marked expansion of tree ferns and scrambling cordaites following breakpoint boundaries (Phillips and Peppers, 1984) and by the massive expansion of tree ferns following the terminal Westphalian extinctions. Tree ferns were cheaply constructed; had massive reproductive output, causing vast overrepresentation in the miospore record (Willard, 1993); produced small and widely dispersed isospores capable of founding a population from one spore; and underwent a major radiation in coal swamps and clastic lowlands during the late Westphalian and Stephanian. Similar patterns have been detected at the Cretaceous-Tertiary boundary, the now famous "fern spike" (Tschudy et al., 1984), and the great radiation of angiosperms in the early Tertiary may have been made possible in part by their fundamentally opportunistic biologies during a period of ecosystem disruption (Wolfe and Upchurch, 1986; Wing and Tiffney, 1987).
A secondary dynamic to result from this process is the evolution of larger, longer-lived, presumably more competitive species from opportunistic ancestors. It is the larger forms that ultimately dominate subsequent ecosystems. The tree ferns of the Stephanian again exhibit this clearly, and hints of similar patterns are found among the pteridosperms, lycopsids, and sphenopsids. Most Westphalian marattialeans of coal swamps were not truly large trees; Lesnikowska (1989) has reconstructed scramblers and small trees in which virtually all measurable indicators of size are smaller than descendant forms in the Stephanian, in some instances up to an order of magnitude smaller. Gigantism in Stephanian coal-swamp plants has been noted for sigillarians, which our studies suggest were much larger than intraswamp sigillarians of the Westphalian. Galtier and Phillips (1985) and Willard and Phillips (1993) note record large sizes for medullosan and sphenopsid stems in Stephanian peat deposits. Large sizes point to longer-lived plants in Stephanian coal swamps, and taxonomy points to a new suite of species. Together, these observations suggest that Stephanian descendant forms, which had become ecosystem dominants, were growing in more long-persistent habitats than were Westphalian forms, which had been largely subdominant to lycopsids or ecologically restricted by them during the Westphalian.
The fabric created by the biotic interactions among organisms plays a large role in dictating evolutionary dynamics. Different kinds of opportunities for establishment of divergent phenotypes appear to exist during times of environmental stability than during times of instability and disruption. This is strongly suggested by species turnover, both its magnitude and the ecomorphic nature of species replacement. We suggest that during periods of environmental stability the targets of opportunity for divergent forms are strictly defined by both biotic and abiotic factors. Occupied niches are for the most part not available to new phenotypes, creating a very fine-meshed selective filter. In contrast, during times of environmental change, and some degree of consequent disruption of the