New data and novel analyses invigorate old debates. A provocative example, published in the journal Ecology, comes in the form of a paper by Kelly et al.1 that will spark fresh argument over the question of the factors that determine patterns of biodiversity.

Traditional explanations for the local co-existence of species hold that the balance of nature is delicately related to differences in how species interact with their local environments (their 'niches'), with populations of each species being primarily regulated by distinct environmental factors2. Such niche partitioning results in stable frequency dependence, in which each species increases relative to others when it is rare, and decreases when it is common3. This venerable view has been confronted with the contention4,5, arising from recent modelling work, that stochastic demography and dispersal are more important, and that they allow the widespread coexistence of species with identical niches. This 'neutral theory' has provided possible explanations for the occurrence of highly diverse communities that challenge the traditional view, and has indicated ways to account for them with simple models.

The resulting clash of ideas has led to a possible synthesis that finds a place for both niche and neutrality6,7. The key to this synthesis is the thought that niche partitioning is increasingly less likely, and neutral dynamics more likely, as species come to resemble each other more closely. This synthesis would retain niche partitioning as a component of community structure, but would use neutral theory to resolve the matter of why niche partitioning cannot by itself explain highly diverse communities.

This is the context in which Kelly and colleagues' work1 is set. They examined abundance patterns of trees in a highly diverse tropical forest in Mexico, and focused especially on pairwise patterns of relative abundance between closely related (congeneric) species (Fig. 1, overleaf) and less closely related species. They found that coexisting congeners had relative abundances that indicated stable frequency dependence, whereas less closely related pairs of species showed patterns consistent with neutral theory. These results are surprising — not only is it the most closely related pairs of species that show frequency dependence, rather than less closely related pairs, but also the less closely related pairs show relative abundances similar to those expected with neutral dynamics. Both of these aspects contradict the proposed synthesis described above. The authors also show that the type of frequency dependence manifested by closely related pairs of species is consistent with, and indicative of, mechanisms associated with temporally fluctuating environments. This suggests that such species-pairs coexist because one does better than another at different times, for instance during periods of drought, rather than because they specialize on different resources.

Figure 1: Common ground.
figure 1

C. K. KELLY

Two adult congeneric trees — Bursera instabilis and B. heteresthes — in intimate coexistence at Chamela Biological Station in Mexico, the site studied by Kelly et al.1. Although the two species are distinct in bark coloration, they are otherwise similar in most respects.

Full size image

At first glance, these results imply that closely related species show niche partitioning, whereas interactions between less closely related species are determined by neutral dynamics. However, this raises two puzzling questions.

First, how similar do species have to be to result in neutrality between them? Kelly and colleagues' findings indicate that, even in a highly diverse community, closely related congeners are not similar enough to show neutral dynamics. One possibility is that close relatedness is not reflected in ecological similarity. But this seems unlikely, because such relatedness is highly correlated with similarity of growth form, demography and general ecology.

Second, to what degree does finding relative abundance patterns consistent with neutrality really indicate that stochastic demography and dispersal regulate interactions between species-pairs? A notable point here is that Kelly et al. looked at pairwise relative abundance distributions, whereas most previous studies examined such distributions in entire, and often highly diverse, assemblages. Kelly and colleagues' findings tend to confirm that the fit of relative abundance distributions to neutral theory is a poor diagnostic for the importance of stochastic demography and dispersal. It may instead be that the structure of environmental variation among factors in the community itself may have random components. Some previous models of niche partitioning8,9 have such features, and to some extent these models match the data of Kelly and colleagues.

What is surprising is that this component of niche partitioning does not also affect closely related species. The authors argue that the interactions between such species differ in type from those involving less related species because they show the signature of temporally fluctuating environments, whereas less closely related species seem less likely to be affected by this kind of niche partitioning. The mechanisms that generate temporal fluctuations in niche relations are not known, but such fluctuations may account for some aspects of highly diverse communities. However, it still seems unlikely that they can fully explain the apparent coexistence of hundreds of species in some communities, including this one.

Thus Kelly and colleagues' results provide answers to some aspects of the debate between neutralists and niche partitioners. But they leave open other issues that the tentative synthesis seemed to have resolved.

One issue is whether there is a role for neutral dynamics at all. The new findings provide surprisingly little support for such dynamics but do not rule them out. Perhaps ecological neutrality does occur among some subsets of species in these communities, but, if so, it is not closely associated with phylogenetic relatedness. The other issue is the need for an explanation for the high diversity in these communities. The appealing component of the synthetic view was that stochastic demography and dispersal could enhance diversity in a community with limited niches to an arbitrary degree that would depend on external factors (speciation and biotic exchanges, for example). If stochastic demography and dispersal are not important, even in a diverse community, explanations for diversity must be found elsewhere.

Kelly and colleagues' findings1 indicate that much of the unresolved component of diversity involves the co-occurrence of comparatively unrelated species. This is a clue that the processes involved may not fall in the normal realm of conventional niche theory. This theory has focused on local dynamics, but regional factors may affect local patterns in unpredictable ways. For example, many species may be present as 'sink populations' that are maintained by dispersal from other locations where they have higher fitness as a result of different local conditions10. This is just one possibility, but it does suggest that regional factors may be a key to understanding highly diverse communities such as these tropical forests.

Where does all this leave us? My view is that Kelly and colleagues' paper will turn out to have a considerable impact — but as much in stimulating fresh thinking as in directly clarifying the relative roles of neutral and niche-partitioning processes.