Species pools and the fallacy of composition (UPDATED)

A species pool is the set of all species that could potentially colonize some local site. Many important ideas in community ecology and biogeography start from assumptions about the species pool, and about dispersal from the species pool to the local site, and from those assumptions derive the predicted consequences for properties of the ecological community occupying the local site. MacArthur and Wilson’s theory of island biogeography is the classic example. Other examples include studies of the relationship between local and regional species richness, Hubbell’s neutral theory and modifications thereof, and almost any study of local community structure which asks whether the species occupying a local site comprise a non-random subset of some larger pool of species (e.g., non-random with respect to their phylogenetic relatedness, or with respect to their phenotypic traits).

The notion of a species pool is valuable in calling attention to the fact that local communities aren’t closed systems, and that the species occupying any local site typically came from somewhere else. And if your main focus is on what happens locally, it’s natural to just take “somewhere else” as a given, and call it “the species pool”.

But just because this move is natural doesn’t make it right. Not that it’s necessarily wrong, either. But at best, any explanation of local community structure which appeals to the properties of the species pool is incomplete. Not because it leaves the properties of the species pool itself unexplained (every scientific explanation takes something for granted). But because, in any complete explanation of local community structure, the properties of the species pool won’t be exogenously determined. There’s nowhere on earth that’s external to community ecology; every population of every species is and always has been part of some local community somewhere. Species abundances, always and everywhere, are determined by birth, death, and dispersal: it’s “community ecology all the way down”. So a complete, ultimate explanation for local community structure wouldn’t involve an external species pool at all, or if it did the “species pool” would simply be the sum of the species present at all the localities. The species pool would be a dependent rather than independent variable, emerging as the aggregate outcome of local scale processes of demography and dispersal within and among many local sites.

Appeals to properties of the “species pool” as an ultimate explanation for local community structure suffer from the fallacy of composition. This is the fallacy of thinking that something which is true of some part of a whole must (or even can) be true of the whole. For instance, in economics, an individual who wants to accumulate more savings can do so by cutting back on his or her spending. But if everyone tries to do this at the same time, the result is a recession (the “paradox of thrift“): in the limit, if no one spends money on anything, the entire economy grinds to a halt. As a second economic example, an individual nation can attempt to grow its economy by devaluing its currency in order to make its exports cheaper. But it’s impossible for every nation to simultaneously devalue its currency relative to that of every other nation, or for every nation to be a net exporter.

The fallacy of composition is especially tempting when any given part of the whole is only one of many. Any given local community is a small part of the world, and might receive colonists from a very large area. And the collective dynamics of all the local communities in that large area are not much affected by what happens in any one locality within that area. But just because regional dynamics are independent of what happens in any one locality does not mean they are independent of what happens in all localities. To think otherwise is to commit the fallacy of composition.

The issue I’m raising may seem obvious, and I doubt any ecologist would actually deny what I’ve written above. But in practice, I think we’re so used to taking the notion of a “species pool” for granted that it steers our research efforts without us even realizing that the steering is happening. For instance, if you think of the “regional species pool” as exogenous to the local communities within the region, then it’s natural to do things like treat local species richness as a “dependent” variable which can be regressed on an “independent” variable like regional species richness. Which means you probably don’t even ask about how the resulting local-regional richness relationship would behave if, as is actually the case, the “region” was just a bunch of local communities linked by dispersal (a metacommunity). Which means you fail to ask questions about the joint determinants of local and regional richness, and the (epiphenomenological) local-regional richness relationships to which they give rise (e.g., Shurin and Allen 2001).

I’ve been surprised that the recent surge of interest in metacommunities hasn’t lead to a major shift in research effort, away from lines of research which depend on assuming exogenous species pools. I’m not sure why that shift hasn’t happened.

UPDATE: Ace macroecologist Brian McGill pops up in the comments with some sensible pushback. I was going to reply to him at length–when I realized that I’d already done so, months ago! See this old post, which I’d forgotten about–and which the present post basically just rephrases. I find this slightly embarrassing, to be accidentally repeating myself after so short a time. Especially since I like the old version better!

12 thoughts on “Species pools and the fallacy of composition (UPDATED)

  1. To me, both approaches (island biogeography à la MacArthur & Wilson vs. metacommunity ecology) make sense in the same way that statistical mechanics have different set of assumptions to design thermodynamical models: you can force the focal system through matter flow, temperature forcing, etc. It makes sense to study open systems and closed ones, or closed systems in temperature baths vs. closed systems with variable temperature.

    Likewise, I would not argue against studying community ecology using a species pool if e.g. some part of the system is ‘large’ and autonomous, so that it influences the focal, smaller system, but is not influenced much by it. In evolutionary ecology, Bob Holt and Richard Gomulkiewicz popularized the concept of ‘black-hole sinks’ which is an extreme version of this situation.

    The practical argument for using species pools is that it is simple. There is a silver lining, for sure, because a model based on species pools will never bring an unbreakable argument about any issue – but can another approach really bring such a perfect argument?

    • Oh, I’m fine with appropriate uses of species pools–I’ve even published models and experiments based on them. As long as we recognize the limitations of what we can learn from thinking in terms of species pools.

      I think the analogy to open vs. closed systems is a misleading one–the only possibilities are not a closed local community vs. a community open to immigration from an exogenous species pool. A metacommunity as a whole is closed, but the local communities within it are open–specifically, they’re open to other local communities.

      Not sure what you mean by species pools providing an “unbreakable” or “perfect” argument about any issue, or why this would be an advantage of models based on species pools. It certainly is possible to test and reject models based on species pools.

  2. OK, I see where you are heading. If the question is ‘why haven’t we stopped using species pools to explain community properties because such approaches are necessarily incomplete?’ the answer is probably because it is easier and also because we haven’t discovered everything about species pool models.

    By the way, I have seen very few studies that actually claim that species pool models provide “an ultimate explanation for local community structure”. In most cases, authors acknowledge the fact that species pools have to been explained in some way.

  3. Jeremy – I agree with about 90% of your opinions, but I’ve got to disagree with this. First – just because you CAN reduce the regional pool to the outcome of thousands (millions?) of local species interactions (“it’s “community ecology all the way down”) doesn’t mean you SHOULD. If you’re a chemical engineer charged with making a vat-sized chemical reaction go faster, you don’t use quantum mechanics, or even molecule-by-molecule simulations. You use laws at a useful level pertaining to enzyme kinetics, etc.

    Second, if you really fully buy into this reductionist program, why stop at community ecology. Why not reduce everything to physiology? (every birth or death is just an outcome of physiological processess …) or quantum mechanics? How far down to go is really nothing more than an aesthetic preference or (chauvinism for ones own field?).

    Third, I don’t even agree that you can take a concept like regional species pool down to community ecology in a meaningful way. The model of thousands of species interactions is so hopelessly unparameterizable and chaotic that it is useless. If we use a more simplified model, then we’re really just choose what level of emergent phenomenon we find useful. And as Francois noted, species pools have proven to be a rather useful emergent phenomenon.

    I’m quite happy with starting with a climate-species pool richness correlations to explain regional diversity. I have little hope we’ll ever explain species richness at regional scales by bottom-up community ecology processes (the one great attempt – intermediate disturbance hypothesis – is problematic as you’ve noted). Personally, I think bottom-up community ecology approaches are useful at some (smaller) scales and regional-pool approaches are useful at some (larger) scales.

    • Hi Brian,

      Always happy to get pushback, even if it’s only 10% of the time. 😉

      I actually have an old post which anticipates at least some of your comments, so I’ll just link to that rather than recapitulate it: http://oikosjournal.wordpress.com/2011/05/03/why-macroecology-needs-microecology-revisiting-an-oikos-classic/

      In general, I’m happy to agree that a “macroecological” perspective can be very fruitful, and that trying to do molecule-by-molecule simulations of a chemical reaction vessel (or the “microecological” equivalent) is usually some combination of impossible and silly.

      But trying to infer microecology from macroecology often leads to serious mistakes, especially when (in contrast to the case of simple chemical reaction vessels) we don’t know how to scale from the microscale to the macroscale. Think of trying to figure out if communities are “invasion resistant” or if “species interactions are strong” just by looking at plots of local vs. regional richness, for instance. That inference is *highly* problematic, quite apart from technical issues like appropriately defining the regional species pool associated with any given local site.

      I freely admit that we usually don’t know how to take a bottom-up approach and scale from the microscale to the macroscale (although I’d also argue that there are exceptions, and that these exceptions are among the most successful examples of macroecology; see the old linked post). But the fact that we usually don’t know how to scale from micro to macro doesn’t mean it’s correct to think of the macroscale as somehow free-floating, as if it was completely independent of (rather than a complex “emergent” outcome of) microscale processes. It’s that conceptual mistake that I’m referring when I (somewhat loosely) refer to the fallacy of composition. As I tried to suggest in the post (probably not clearly enough), pursuing a “bottom up” approach can mean much more than trying to do the impossible and parameterize an individual-based model of an entire ecosystem.

  4. Pingback: Are feedback loops ubiquitous in ecology and evolution? | Dynamic Ecology

  5. Pingback: Scaling up is hard to do: a response | Dynamic Ecology

  6. Pingback: Why, and how, to do statistics (it’s probably not why and how you think) | Dynamic Ecology

  7. Pingback: Zombie ideas in ecology: the local-regional richness relationship | Dynamic Ecology

  8. Pingback: Friday links: good cartoons of bad arguments, and more | Dynamic Ecology

  9. Pingback: Book review: Community Ecology by Gary Mittelbach, and Community Ecology by Peter Morin | Dynamic Ecology

  10. Pingback: Steven Frank on how to explain biological patterns | Dynamic Ecology

Leave a Comment

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.