Book review: The Theory of Ecological Communities by Mark Vellend

j10914

Earlier this fall I read Mark Vellend’s The Theory of Ecological Communities. I read it on my own, and also read it in a reading group with several ecology grad students. Here’s my review.*

tl;dr: It’s a very good book that fills a real pedagogical need. Whether it will also shape the direction of future research in community ecology is an open question, I think. Below the fold you’ll find me engaging with the book, which I think and hope Mark will welcome.

The book expands on Mark’s 2010 paper arguing that what Mark calls “horizontal” community ecology (those bits of community ecology concerned with explaining  coexistence, diversity, species composition, and abundance within trophic levels) is closely analogous to evolutionary biology. There are only four “high level” forces at work in horizontal community ecology, each of which is analogous to one of the four “high level” forces at work in evolution: selection, drift, migration, and mutation. Community ecologists mostly haven’t recognized this because for historical reasons they’ve focused their theoretical and empirical work at a much “lower” level. For instance, the classic MacArthur-Levins model of coexistence via resource partitioning merely describes one low-level mechanism, among many others, by which “selection” among competing species can be negatively frequency-dependent and so promote stable coexistence. Recognizing this analogy to the four forces of evolutionary biology helps unify and organize what otherwise seems to be a polyglot collection of unrelated special cases in community ecology. Mark argues that this unification has considerable pedagogical benefit to students, by giving them a “road map” of the field. He also argues that his proposed unification should shape future research in the field. He illustrates his argument by reviewing essentially all of the horizontal community ecology literature, showing how it all fits within his proposed framework and summarizing its major conclusions and key gaps in terms of his proposed framework (e.g., what are the most important sources of “negatively frequency-dependent selection” in community ecology?)

Mark says in his introduction that he set out to write the book he wishes he could’ve read as a graduate student. I think he succeeded. The tables alone were worth the price of admission in the eyes of the grad students with whom I read the book, only one of whom is a community ecologist. To highlight just one of the many tables: Table 5.1 lists essentially every major theory/model/idea about species coexistence and diversity (R* theory, neutral theory, Janzen-Connell effect, island biogeography, IDH, spatial storage effect…), summarizing each in terms of both the key low-level and high-level processes, and giving key citations. The students loved the tables so much they wanted to see them used to open the chapters, rather than as chapter summaries.

Having said that, the book isn’t entirely successful in its pedagogical goals. My sense from the reading group is that students who don’t already have some background in community ecology often will struggle to make sense of Mark’s many 1-2 sentence descriptions of this or that theoretical model, idea, or empirical research approach. I was completely fine with Mark’s brevity–but of course, I’m an experienced community ecologist. I think the book will be most useful to students who are going to work on horizontal community ecology. They’ll either already have, or will go on to obtain, sufficient background knowledge to mentally fill in the many details Mark glosses over. A non-community ecologist just looking for a book-length overview of community ecology might find that Mark’s book assumes too much familiarity with the field.

I’m still a little unsure about the value of Mark’s book to researchers as opposed to students. The conceptual unification it provides is intellectually satisfying to me, and I’m sure to many others. And  Mark’s framework provides the basis for a lot of new empirical research. For instance, now that Mark’s shown how concepts like “selection” apply in community ecology, others are in a position to go out and quantify those concepts in different communities to see if any empirical patterns emerge (e.g., Fox et al. 2010). But I’m unsure how much Mark’s book has to offer someone who says, “I work on low-level process X. I don’t care if I that’s just one of many low-level processes giving rise to high-level process Y.” Or what Mark’s book has to offer someone who says, “I want to explain pattern Z in terms of low-level processes. Knowing which high-level processes those low-level processes give rise to doesn’t help me explain pattern Z in the way I want to explain it.” This is one illustration of a broader point that Mark himself has made in comments here: different scientists have different motivations and goals. And it’s difficult to show that somebody who doesn’t care about goal X is wrong not to care. Perhaps one way to convince that hypothetical someone would be to talk more about what mistakes he or she might well be making by his or her own lights by maintaining an exclusively “low level” focus. I had a go at this in an old post on mistakes “low level” metacommunity ecology could avoid by taking a page from “high level” population genetics. In general, people who think exclusively in terms of “low level” processes are prone to certain sorts of errors (Bill Wimsatt is good on this). And people who think exclusively in terms of “high-level” processes are prone to different sorts of errors. For this reason, I think it’s best to go back and forth between “low level” and “high level” thinking. Each helps guard against the errors into which the other tends to lead you. Mark’s high level framework is thus a complement to, not a substitute for, the low level ideas that it unifies and organizes.

Implicit in Mark’s review of the empirical literature is a particular point of view on “generality” in ecology, one I think Mark shares with many community ecologists. The sort of generality Mark cares about is basically “vote counting”. What he really wants to know is often X happens in nature, how large the effect of high level process Y typically is, etc. So when reviewing the empirical literature, he seems most impressed by, and seems to place most weight on, research approaches that “scale”: approaches that can be and have been applied often, in many different systems. He certainly respects, say, research that takes advantage of the unique features of some model system in order to conclusively demonstrate the importance of some high level process in that system. But at the end of the day he doesn’t consider work in model systems as a sufficient basis for community ecology. His whole book is about the importance of #4 on my list of “roads to generality” in ecology–but only as a means to the end of #1 on my list. (Personally, while I can appreciate the value of “vote counting” generality, I care more about other senses of “generality”).

One side effect of this, I think, is Mark isn’t always quite as critical as I might’ve wished about the flaws and limitations of certain popular empirical research approaches. If you want to be able to do community ecology “at scale”, you can’t afford to be too choosy about your research approaches. For instance, you’re going to be open to any broadly-applicable, “off the shelf” approach that purportedly lets us infer process from pattern, even though such approaches have a lousy track record in community ecology. Because the alternative is to be left without the ability to say much of anything about what’s going on in many communities. Better an untrustworthy but broadly-applicable answer than no broadly-applicable answer at all, I can imagine Mark (and many other community ecologists) saying. Whereas personally, I’d rather have good answers from a smaller number of systems than bad or untrustworthy answers from a larger number of systems. Mark casts a wide net in reviewing the empirical literature. He reviews the results of every relevant line of empirical work from the recent and recent-ish literature. He notes the major criticisms and limitations of each approach (e.g., of local-regional richness relationships as an inferential tool), but rarely goes beyond noting them. He clearly wanted his book to reflect the literature rather than criticize the literature. Which probably just shows that someone like Mark is better-suited than someone like me to write a book like this. Mark is the most broad-minded, fair, curious ecologist I know. You probably wouldn’t want someone as opinionated and skeptical as me writing an entry point into the literature.

So those are my main thoughts. But one sign of a good book is the sheer number of interesting thoughts it prompts in the reader. By that measure, Mark’s book is very good, because I have many other thoughts about it:

  • The style is mostly clear and straightforward without being distinctive. Except that occasionally Mark switches to a more personal voice, as when he complains about the number of papers published annually on horizontal community ecology (“TEN THOUSAND PAPERS!”). I found this funny and charming. To write this book, Mark took it upon himself to

read-all-the-things

and it sounds like he found the experience exhausting!

  • I found myself wishing–greedily, I know–that Mark had devoted a chapter or even multi-chapter section to reviewing how evolutionary biologists have developed and tested high level theory about selection, drift, mutation, and migration. Perhaps there would be some lessons for community ecologists? But that would’ve been a very different book.
  • As Mark notes early in the book, he’s basically redrawing the boundaries of community ecology so as to exclude much of what’s traditionally been considered part of the field. Everything to do with “vertical” interactions like predation and parasitism, for instance, except insofar as they affect species coexistence. I like that redrawing. But it’s definitely a debatable point. Do you think that community ecology should be split into two separate subfields with only modest overlap–Mark’s “horizontal” community ecology and “vertical” community ecology?
  • Mark’s Fig. 4.3 shows how the many “low level”, system-specific causes of fitness differences give rise to a few “high level” consequences for selection, which in turn has a few universally generalizable consequences for community patterns. I love, love, love this figure, because you can draw the analogous figure for other broad areas of ecology. For instance, I draw exactly the same figure in my population ecology class. I learned it from Ed McCauley, who used to teach the class. Ed and I teach that the innumerable and often system-specific “low level” processes affecting population dynamics can be aggregated into a much smaller number of much more general “high level” processes, which Ed calls “dynamical mechanisms”. The full list of “high level” processes in population ecology: intrinsic growth rate, density dependence, population structure, environmental and demographic stochasticity, time lags, nonlinearities, and dispersal. From those higher-level processes result a relatively limited number of different sorts of population dynamics. So it’s not just that community ecology has lacked the high level unifying framework that evolutionary biology has. Other areas of ecology have the high level unifying framework community ecology has lacked. (As an aside, this is why John Lawton was wrong when he famously argued that we should give on doing community ecology because population ecology has general principles whereas community ecology is a stamp collection of special cases. John was comparing a “high level” picture of population ecology with a “low level” picture of community ecology, which is an apples-to-oranges comparison.)
  • I disagree with Mark’s suggestion that students who want to get started with modeling should begin by coding up simulation models, rather than by trying to learn a bit of math. I think you’re better off learning some simple mathematical models first. Or at least first coding up some simple mathematical models rather than whatever complicated scenario you ultimately want to model. I think if you’ve learned a bit of math, you’ll have learned to think in terms of rates rather than amounts, and so better be able to avoid simulating a scenario that doesn’t actually make sense, or misinterpreting the reason why your own simulations behave as they do. I don’t think you can learn to think in terms of rates very well by just coding, though I’m open to debate on that. In fairness, Mark himself uses  his own simulations to illustrate various aspects of model behavior, all of which are already known from mathematical analyses of those same models. That is, Mark uses the simulations to illustrate the math, which I think is fine.**
  • The end of the book has some astute suggestions for future research directions. I love the suggestion of experimentally manipulating effective community size, and thus drift.*** I actually did that myself a couple of years ago–and haven’t published it yet because, as Mark notes, it turns out to be really hard to manipulate effective community size without changing the strength of selection as a side effect. Even in laboratory microcosms. 😦 I also agree with Mark’s remark that modern coexistence theory isn’t (yet) a theory of species diversity; how species coexist is a different question than how many species coexist. And I agree with Mark that questions about the relative importance of different “high level” processes are natural ones to ask, while remaining a bit skeptical that natural questions are always good ones. See for instance this old post questioning whether it makes sense to ask about the relative “importance” of the ecological analogues of drift vs. selection at the whole-community level, as opposed to at the level of individual species within communities.
  • There’s a technical mistake. In section 6.3.3 and Fig. 6.4, Mark incorrectly says that fluctuating selection is an equalizing mechanism that slows competitive exclusion between two competing species. This is incorrect because he’s implicitly comparing to the wrong control (Fox 2013). He’s comparing a case in which selection constantly favors one species over the other to a case in which selection does not favor either species on average, and also fluctuates around the average (sometimes favoring one species, sometimes the other). Mark, like many other ecologists, is thus confounding fluctuations in selection with changes in the average selection coefficient. The correct control to isolate the effects of fluctuating selection is a case with constant selection of the same long-term average strength and direction as in the fluctuating case, but no fluctuations around the average. Comparing to the correct control shows that fluctuating selection of the sort illustrated in Fig. 6.4 doesn’t promote coexistence, not even as an equalizing mechanism (Chesson & Huntly 1997, Fox 2013). From the way section 6.3.3 is written, I’m guessing Mark is aware of the issue here, but sees it not as a mistake but rather as a technicality he lacked the space to get into. If so, I respectfully but strongly disagree with that choice. Indeed, I think Mark himself disagrees with it! Because when he discusses theoretical and empirical work on spatially-varying selection, he’s explicit about the importance of not confounding the spatially-averaged strength or direction of selection with variation around the spatial average. The exact same point could have and should have been made in the context of temporally-varying selection, without taking any extra space. To not do so perpetuates–or at least misses an opportunity to slay–a zombie idea about how fluctuating selection affects species coexistence. But this is a hangup of mine, so I would say that. 🙂
  • Finally, I was surprised and flattered to be cited more than once in the book. Mark even cited one of my blog posts! This takes me back, because one of Mark’s first contributions to blogging was to dig up information on how to properly cite blog posts. 🙂

*Full disclosure: Mark is a friend. He’s written a couple of guest posts for us about issues raised by his book. He was kind enough to send me an author’s copy of his book. I gave him a bit of feedback on part of the book, for which I’m acknowledged. Perhaps most importantly, I’m very much on Mark’s wavelength. I would’ve reviewed the book even if it had been written by a stranger; the subject is right up my alley. And I’d like to think that my friendship with Mark makes it easier for me to be honest. I know I can trust Mark not to be offended by anything I say. But I read the book as someone who sees the field of community ecology more or less as Mark does (e.g., this and this). So I’m not part of the primary audience for the book, and am not well-placed to evaluate how it would come across to someone who was coming to it cold, or coming to it inclined to disagree with it.

**Also, contrary to what Mark says, mathematical modeling is mostly not about finding closed form analytical solutions. So intractability of a closed-form analytical solution is not a reason to skip straight to coding up numerical simulations. For instance, you can still learn a lot about the behavior of a model lacking a closed form analytical solution just by (e.g.) using algebra to solve for any equilibria, using calculus to check their local stability, etc. Even if you’re talking about a stochastic model, you can still use math to derive key properties of the stochastic process.

***In part because the existing evidence that drift matters sure is weak. I leave it to you to decide how depressed/embarrassed we all ought to be about this, given that drift has been a hot topic for 15 years. Perhaps some of the effort that has gone into showing that this or that empirical pattern is consistent with drift (well, drift plus absence of selection) should’ve gone into severe tests of drift. Related.

20 thoughts on “Book review: The Theory of Ecological Communities by Mark Vellend

  1. Thanks very much for this Jeremy. From your comments and others, I have some confidence that students will find the book pedagogically useful. Only time will tell whether it influences research in a major way. When I was trying to get the paper published, I made an effort to convince people ‘yes it will’, but really that was never the primary goal, although of course that outcome would be gratifying. I’ve come to realize that when the premise of a paper/book is one key idea, it’s really not possible to have a “goal” for it at all: the idea arises, and you see where it leads you. The book was like an experiment to see how far it can be taken, and I look forward to seeing whether/how it influences people’s thinking, teaching, and research.

    “rather have good answers from a smaller number of systems than bad or untrustworthy answers from a larger number of systems”. I don’t think anyone wants bad and untrustworthy answers about anything. Reality is that we probably have decent and imperfect answers from many systems AND good ones for a few. Seems reasonable.

    “wanted his book to reflect the literature rather than criticize the literature”. Indeed. Criticizing every corner of such a large literature would depressing (albeit important).

    Math vs. simulations. Not sure I have a perspective on what people “should” do, but rather a perspective as a teacher that I can get a classroom of students going on simulations very quickly that let them explore for themselves how selection, dispersal, and drift interact. Anyone who really wants to delve into ecological theory should certainly learn math. For everyone else, I think simulations are a fantastic beginner’s learning tool.

    “incorrectly says that fluctuating selection is an equalizing mechanism that slows competitive exclusion between two competing species…sees it not as a mistake but rather as a technicality he lacked the space to get into”. I tried to sort out the different perspectives on this, seeing validity in arguments on both sides. Give me some kind empirical implementation that can be explained without causing a headache, and I’ll use it!

    Thanks again, Jeremy, and to others who have provided feedback along the way…

    • “I’ve come to realize that when the premise of a paper/book is one key idea, it’s really not possible to have a “goal” for it at all: the idea arises, and you see where it leads you. The book was like an experiment to see how far it can be taken”

      Very interesting.

      ” I don’t think anyone wants bad and untrustworthy answers about anything. ”

      True. I probably put that badly. What I meant is that some people are more or less ok with known flaws or limitations of approach X, on the grounds that no approach is perfect, that at least approach X is broadly applicable. They may also tend to hope that approach X will be improved in future.

      EDIT (sorry, hit “post” before I was done):

      “I tried to sort out the different perspectives on this, seeing validity in arguments on both sides. ”

      I remain really puzzled why you think there are two sides to sort out here. Do you not see the case of fluctuating selection in time as analogous to spatially-variable selection in space? Why do you see “two sides” in the case of temporal variation but not in the case of spatial variation? And if the answer is that different empirical studies have been done on spatially- and temporally-variable selection, well, I’m afraid I don’t understand why you see the issue as empirical rather than conceptual. If an experimental design is confounded, it’s confounded, even if nobody in history has ever done the unconfounded experiment. I mean, imagine a world in which no one had ever run an experiment with spatially-variable selection (or, say, spatial variation in some environmental variable) compared to a control treatment with the same average strength of selection and no spatial variation (or, say, the same average value of the environmental variable with no spatial variation). People had only ever run experiments that confound spatial variation in selection (or spatial variation in the value of the environmental variable) with the spatially-averaged strength of selection (or the spatially-averaged value of the environmental variable). In that hypothetical world, would you say there are two sides to the question of whether or not existing studies confound the spatially-averaged strength of selection with variation around the average?

      As for an empirical implementation that could be explained (and implemented) without a headache, what’s so hard about that? It’d be trivially easy to run a microcosm experiment that compares fluctuations in some some environmental variable with a control treatment holding that variable at its temporally-averaged value. In fact, I’m sure that experiment’s been done. And it only took me one sentence to describe the design.

      I’m sorry (seriously), I must sound obnoxious or crazy to still be hung up on this. But I just don’t get where you’re coming from on this, and I don’t like that. I really want to get where you’re coming from, so that I can (hopefully!) then address it. Because if I can’t get this point across to you, I have no chance of getting it across to anyone, and I’ve utterly failed as an explainer.

      • I do believe I understand the mathematical argument – so no explainer failure. Let me come at it from another angle. I think the disagreement in the literature might stem, in part, from the idea of a “correct control”. Imagine a northern forest that never experiences fire. It has a dense canopy of trees and low plant diversity. A second forest (otherwise identical) burns periodically, eliminating none of the species that live in the first forest (they can survive an occasional fire), but allowing colonization and persistence of a bunch of other species (e.g., fireweed). What are the theoretically viable explanations for more diversity in the second forest? What is a “correct control”? (All questions , no answers here, but maybe it’s a start towards figuring things out.)

      • Ok, fair enough. You can argue that, in nature, the comparison we ultimately care about is between (say) “undisturbed” and “periodically disturbed”. In light of that, who cares (well, at least very much) whether the effects of those periodic disturbances are because they raise the long-term *average* mortality rate, or because they introduce temporal *fluctuations* in the mortality rate? As long as we get the *net* effect of changes in the average and changes in the variance right?

        Personally, I do care about the distinction between effects of the average and effects of fluctuations around the average. In part just because I’m the sort of person who cares about this sort of thing–I’m a hair-splitter by nature. But in part because I can imagine real-world empirical cases where there’s an objective reason to care. E.g., because you’re comparing two systems with, say, similar long-term average conditions but different variances around the average. Such cases do exist in nature. So I think any “viable” theory will draw that distinction.

        Also, if your question is “why are there more species in the second forest than the first” as opposed to “how do the species in the second forest manage to coexist?”, well, those are two different questions, as you astutely note near the end of your book. But I think the same point still holds. I think that, to answer to the question about diversity differences, we’re still likely to want to distinguish effects of average conditions from effects of variance around the average.

        I completely agree with you that the vast bulk of the empirical studies in ecology aren’t designed to distinguish effects of introducing temporal fluctuations from effects of changing the average conditions.

  2. Hey Jeremy! You proud of me now? Actually commenting on your blog.

    First I will say that the group of us that read the book at UVic loved it. We too thought it was very useful synthetic oversight of what sometimes feels like a confusing mess of ideas to students like me. I found the insight to group all those different processes into four “high level” process quite genius. It really helped me frame my thinking in ecology. However, we had one major problem with the book. What is proposed is not a theory (or if it is, not a very good or useful, i.e. rigorous, one). It does not explain some confusing aspect of the world or explain things better than what we already had (he just organizes those explanations). One physicist I read says: “A successful theory you get more out of then you put in”, meaning it makes strange predictions or helps you see the world differently than you did before. This book is wonderful at organizing disparate thoughts, but it doesn’t generate any really interesting predictions to test (other than perhaps those in the final chapters). In fact, the presented predictions cover every base. You could go out and measure anything essentially and it would fit somewhere under one of the predictions; you observe a community and if the pattern does not fit negative frequency dependent selection explanations, then it must have been positive frequency dependent selection or drift etc and the theory still holds. As such this theory is not really falsifiable. The main problem being, what I think, are weak hypotheses. Saying something is “important” is not all that useful or falsifiable. Not many ecologists these days would argue those four process are not important. Similarly, I cannot tell how the predictions developed from the hypotheses. To me, they read like they came from the findings of the literature, not from playing around with the theory.

    Despite all this I still LOVED the book. I just wish it had a different name. It is useful as a modern synthesis of community ecology, which will lead to interesting theoretical developments (and perhaps a more rigorous/mathematical theory of ecological communities). Mark had excellent insight identifying the “first principles” of ecological communities (selection, drift, dispersal, speciation) from which a theory, quantitatively describing how selection and drift interact for instance to produce a predictable community, can develop. Maybe Mark can comment as to why “theory” rather than something like “modern synthesis” was chosen as the title and framework for the book?

    Am I reading too much into the name? I just don’t think we should throw terms like “theory” around. Maybe I’m wrong and this is the best kind of theory ecology can produce. But I hope I’m not. I also hope I don’t come off sounding like a brat. Those of us at UVic commend Mark on a tremendous synthetic work really useful for students in particular; we just think calling it a theory may be getting ahead of ourselves. Anyway, that’s just one more opinion!

    • Hi Geoff,

      Better late than never on the commenting. 🙂 And thanks for sharing a grad student perspective.

      I agree with you that the “theory” Mark proposes could perhaps be termed a “conceptual framework” or some other suitable term.

      The question you raise is what the point of such frameworks/theories/whatevers is, besides making it easier for students to organize the literature in their own heads. Afraid I don’t have much more to say on that besides what I said in the post, and in the old posts of mine that I linked to (e.g., the one on what metacommunity theory can learn from population genetics).

      Possibly, recognizing that evolution and horizontal community ecology share the same conceptual framework opens up new opportunities for comparative work spanning both fields. I have an old post suggesting this: https://dynamicecology.wordpress.com/2011/06/24/free-idea-for-a-provocative-review-paper/

    • Thanks for these comments – much appreciated. I have often wondered about how we use the word “theory”, and in looking both at definitions from dictionaries and philosophers of science, and also common usage among scientists, it’s actually quite heterogeneous. Scheiner & Willig’s book (“The theory of ecology”) has an interesting take on things. My short response is that I’m not too concerned with what the book is called – so feel free to paste a different label on the cover!

      As a longer response, some thoughts here on what theory means. What is “modern coexistence theory”? Really it’s a perspective on existing models that shows how they share certain unifying features. What is “metacommunity theory”? Mostly a label that brings together models that already existed and other minor variants in development. What is “evolutionary theory” or “population genetics theory”? When first developed they certainly made non-obvious predictions (“explain things better than we already had”), but what if no one thought to use the word theory until 100 year later when the explanations had been established as obvious? In my case, I apply the word “theory” to the umbrella that unites a great many models and specific hypotheses. Had it been developed and applied 100 years ago, I doubt there would be objection to the word theory. So labelling it as such after the fact doesn’t make it less of a theory, in my view. In other words, you don’t need to call something a theory at the moment it offers an improvement on older explanations for it to still be a theory later. But as I said, I’m fine with other descriptors.

  3. I’m organizing a reading group of the book here at Michigan next semester! One thing I’m debating is whether we should read the 2010 paper before launching into the book. Thoughts on that? Would it work as a good intro, or would it feel redundant?

  4. Inspired mostly by your and Margaret’s posts, I bought this book and I’m finding it awesome. It also gave me ideas for my population ecology course. So thank you for writing about it!

  5. Pingback: Poll: do you think species-rich communities are those with stronger coexistence mechanisms? | 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 )

Google+ photo

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

Connecting to %s