Jordi Bascompte’s “Stucture and dynamics of ecological networks” (Science, 13 August 2010) cites Robert May’s theoretical work in the 1970s as showing that, because “the more complex a randomly built food web, the less stable it is… real networks must have some contrasting, nonrandom structures that allow them to persist despite their complexity.”  Efforts to characterise the nonrandom structures, including Bascompte’s, would benefit from giving more attention to lesser-known theoretical work from the 1970s and 80s that showed that, whereas stable systems may be extremely rare as a fraction of the systems being sampled (May’s result), they can be readily constructed over time by the addition of populations from a pool of populations or by elimination of populations from systems not at a steady state.  Under such a “constructionist” perspective: stable complex ecosystems need not be weakly interlinked modules of populations—they can be more richly interactive; the range of mathematical possibilities that modelers can consider is extended; persistence of complexity does not necessarily require devious strategies (ongoing turnover of populations may be all that is needed); complexity can be constructed in ecological time without shaping of interactions by natural selection; complexity constructed in ecological time depends on its spatial context; and complexity might be better conceived in terms of intersecting processes, not well-bounded systems.[1] As remarked ten years ago in the context of a debate about whether diversity enhances ecosystem function, the constructionist perspective on ecological complexity is also preferable for any serious consideration of the implications of human interventions within ecosystems.[2]

---

[1] P. J. Taylor, Unruly Complexity (University of Chicago Press, 2005), 3-17.

[2] P. J. Taylor, Science  290 (5489), 51