ECOLOGIST Stephen P. Hubbell describes the behavior of species with a--controversial--mathematical formula.

For a graph depicting Hubbell's theory in action, click here.

What determines how many species live in one place and how abundant they are? Many ecologists will reply: "The resources of the habitat and the strategies that the species use to exploit them." Ecologist Stephen P. Hubbell answers with a mathematical formula, which he described in a book last year titled, "The Unified Neutral Theory of Biodiversity and Biogeography." To his colleagues' consternation, the formula seems to work: for trees, bees, bats, birds, fish, frogs and others--accounting for observed patterns of diversity with just a few parameters, such as rates of birth and migration. You'd think ecologists would be ecstatic. But Hubbell started from a postulate that most consider preposterous: that one tree or one bird is just like any other. His patterns result solely from random fluctuations in births, deaths and the arrival of new species.

But isn't being different the way species dodge head-to-head competition and avoid driving each other extinct? Those differences could be incidental, the formula's success suggests. Yet in a study that Hubbell and colleagues published earlier in 2002 in Science magazine, the theory hit a bump in the road. Hubbell's previous tests had concerned biodiversity either at one locale or collectively within a large region--which ecologists call "alpha" and "gamma" diversity, respectively. This time the test was what ecologists term "beta" diversity: how species demographics vary across a landscape. A careful setting of the parameters enabled the formula to predict changes across short, intermediate or long distances (0 to 0.1 kilometers, 0.2 to 50 km or 50 to 1,400 km) in forests of Panama and the Amazon. But no setting accounted for beta diversity as a whole.

With the theory's first failure, the time seemed ripe to ask Hubbell about his radical assumptions. Scientific American phoned him at his office at the University of Georgia, Athens.

SA: In the beta-diversity study of earlier this year, the parameter you focused on was seed dispersal. Why is that?

SH: Well, we were taking data on the occurrence of tree species at varying distances. If trees couldn't disperse anywhere, seeds would just land under the mother and that tree would replace itself with its own babies. There'd be no probability of sampling the same tree twice in different places. But with dispersal, they [the trees] have a chance to spread elsewhere, and so what you can do with the theory is to estimate what rate of spread is necessary to have the observed probabilities of drawing two trees of the same species from two places at a distance, r, apart.

SA: But no single rate explained the probabilities you observed over all distances. That's a problem for your assumption that one set of parameters describes all trees, isn't it?

SH: It's only a problem for the simplest version of the theory. In other words, the one published in the book is highly simplified. Applied to trees, it [the simplified version of the theory] says that all seeds effectively move a fixed distance--the mean distance--and of course we know that isn't true. You could have a more realistic neutral theory, in which lots of the seeds fall right beneath the mother, a few go farther, and even fewer go farther than that. If every tree species followed the same distribution of dispersal distances, then it could still be neutral.

SA: "Neutral" as in there are no differences among trees that matter?

SH: That's right. And if we knew what the distribution function was for dispersal distances, then the theory could perhaps fit the data.

SA: But is there even such a thing as a distribution function for all trees? In the Amazon don't you have some species relying on bats and others using insects or wind or rodents to disperse seeds? I would think birds disperse seeds differently than ants.

SH: That's correct. And so in principle, one might expect that the theory would fail when you have real differences among species. But there's a body of literature that says there's a trade-off between dispersal and competitiveness. For example, pioneer species of tropical trees are selected [through evolution] for high dispersal rates, because they have to find rare light gaps in the forest, but they're not good competitors. If a good disperser is not a good holder of the site it colonizes, it loses that advantage. And if those two are more or less balanced, then the net effect is potentially no different from neutrality.

SA: So because there's a trade-off, you're saying, it doesn't matter whether trees are neutral? The theory just works as if they were?

SH: That's right.

Look, I think the biggest question to come out of the neutral theory is: "Why does it work so well?" I'm as puzzled as the next person. But one idea is these trade-offs. With trade-offs you essentially equalize the fitnesses among species, so that you don't have to uniquely specify each and every resource and all the interactions to characterize an ecological community. A much smaller and simpler set of rules governs their dynamics.

At the same time I think we need to go back to first principles and ask, "Gee, are species really as specialized as we said they were?" If you think about it, any species that's so specialized it can't opportunistically change when its preferred resources become absent or rare, it's going to go extinct. Maybe the species that are around today are ones that in a pinch can make do with something else as a food source or whatever.

SA: I have to tell you, I ran that idea past a colleague of yours--that maybe species aren't so specialized and ecologists just failed to notice--and all he did was chuckle.

SH: I think it's controversial. There are several predictions of the hyper-specialization view that seem not to fit the data. One is that the removal of species should not result in the increase of other species. But most experiments have shown that the removal of competitors results in an increase in the populations of other species. The other thing is that we hardly ever observe competitive exclusion--one species driving another extinct.

I mean, we know that there are specialized species out there, and the theory won't apply to those cases. But many other species may be much less specialized. For example, we know that many insect species are only found on one host tree species. But whether, you know, insectivorous birds can feed on each other's food--that's very highly likely. We don't really know. And there's been a tendency to put on blinders. We presume species differences explain everything we need to know about ecology--but the theories that assume this haven't done very well.

SA: On the other hand, it seems impossible you could ignore species differences and get things right.

SH: Well, I think many of the patterns we see in communities may be the properties of large systems and average rules that have very little biology in them. Populations are subjected to all sorts of unpredictable forces that we categorize as random--and this may be more a statement of our current ignorance than a statement of nature. But I think the neutral theory works because overall communities are saturated, so that any species that increases in abundance does so at the expense of a collective decrease by the same amount of everybody else.

SA: That's your second postulate, isn't it? "The zero-sum rule"?

SH: Right. I think communities really are more or less zero sum. A lot of evidence points to biological saturation. And the result is that even though there are differences among species and even though populations are impacted by unpredictable forces, we can get away with a simple zero-sum-game as a good approximation of reality.

SA: But for beta diversity the approximation doesn't seems so good. I think you even wrote that beta diversity in Panama was higher than your theory could account for. Do you think beta diversity has a different explanation than alpha and gamma diversity?

SH: I don't know. But probably. Let me just say that the larger the landscape scale on which you examine diversity, the more likely you are to get very different habitats in your sample--hills and ravines and bottom lands and river courses--and the neutral theory at present doesn't encompass any heterogeneity of that sort. That's work that needs to be done. So to the extent that species are specialized to habitat, then the theory will not predict that.

SA: Specialization to habitat? That sounds pretty antithetical to neutrality.

SH: Nevertheless, what I see in the future is there will be mixed models that perhaps have neutral dynamics within habitat patches, but not across habitat patches.

This is just like population genetics, where we have natural selection on one hand and we have genetic drift on the other--and we know both are going on simultaneously. In ecology, we have the analogy that we have competition and predation and all these other things going on, but we also have random fluctuations in population size.

What I'd like people to take a way from all this is that the theory provides us a way of testing when biology is necessary as part of your explanation. I'm not insisting this is the way nature works. I'm saying here is a theory that works surprisingly well with few assumptions, and we should let it do as well as it can and add those things that we have to add when the theory fails. I'm not a neutralist. I believe that we needed a formal neutral theory in order to know when we needed biology to explain something.