The July 2015 issue of Trends in Ecology & Evolution features a really important review article entitled “Selection on stability across ecological scales“. The paper embraces the idea that the stability properties of ecological systems dictate the configuration of extant social groups, interacting species pairs, and overall ecological communities. Lev Ginzburg, my Ph.D. advisor, has been kicking around this idea for decades, and it is exciting to finally see selection on systems as an evolutionary mechanism explored in a prominent publication like TREE.
Oddly enough, the article is breaking new ground. Although the idea has been around for a long time, most people do not think of systems being subject to selection. This review nicely lays out the basic concept, which is that that the configuration of systems of interaction can influence how likely those systems are to actually be able to exist (what they call feasibility) and how likely those systems that can feasibly exist are to actually remain in existence (what they call stability). A large part of the review is dedicated to differentiating this kind of “systemic selection” from the kind of natural selection all evolutionary biologists agree ‘acts on’ individual organisms. The authors of this review have chosen the term “nonadaptive selection” to describe selection on systems because, as they explain:
Why is this idea of nonadaptive selection on systems radical (if not entirely novel)? Well, for the most part I think that many evolutionary biologists and ecologists are happy to think about social groups, interacting species pairs, and even ecological communities as just scaled up versions of what is happening at the level of the individual. In other words, a lot of researchers believe that we can can explain all of nature based on understanding individual selection alone. There are many reasons to believe otherwise, starting with the inability of individual-level processes to predict patterns at higher scales; the entire field of complexity theory and emergence are now dedicated to understanding these patterns.
The other reason why this sort of evolutionary thinking is not mainstream is because it is difficult to test empirically. I remember back in graduate school thinking to myself during my umpteenth invited speaker seminar talk “man, is all of evolutionary biology about finding clines and hybrid zones for a particular trait?”. I think that we have to face a painful reality: the big results of our field are highly biased towards what is more tractable (in other words, easier). This is especially true in a scientific climate that is so heavily reliant on highly-contested, limited grant money. Our focus on individual-scale adaptive evolution simply reflects what is easier to study, not what is more important. The same problem exists in genetics, where studying the effects of individual genes trumps the more valuable enterprise of understanding how gene networks interact with environmental factors to produce phenotype.
This review takes on the can you test this kind of theory? question, citing a number of studies at the levels of species pairs or simple food webs that demonstrate that theoretically-predicted patterns consistent with systemic selection can be found in nature (see Figures 1, 2, and 3). The problem is that this is very soft hypothesis testing. What would be more interesting — but more difficult to undertake — would be to show that actual systems are destabilized by the addition or subtraction of species. Unfortunately, this can be hard to do (working with whole systems as your unit of replication is always more difficult than working with individuals as your unit of replication), unethical to do (as destabilizing natural systems would be instructive but also very destructive), and might only meaningfully happen on time scales that are too long for us to study (which might explain the dominance of so-called “adaptive evolution” in our field — at least you can observe it happening). Clearly there is a lot more work to be done in this area, and having a paper like this to cite in a grant proposal will potentially ‘normalize’ this kind of research.
As much as I am excited about the publication of this paper, I am also a bit underwhelmed by it. Having spent a lot of time back in graduate school processing some of these ideas with Lev, I know that there is a lot more to this story that is not being brought to light in this paper.
The first thing that still amazes me — but does not surprise me — is the complete avoidance of multilevel selection in this paper. Admittedly, the kind of ‘selection’ involved in this so-called nonadaptive evolution is not the traditional form envisioned by multilevel selection. But the central value of multilevel selection is something that this ‘new’ theory desperately needs to consider: how do factors influencing survival at different levels of organization contribute to the overall evolutionary process? This paper ignores that multilevel selection exists, presumably not because the authors are ignorant of multilevel selection but because they want to avoid the controversial baggage it brings with it. Check out this important line from the paper:
In other words, we need to look at selection simultaneously operating at multiple levels. If you need more convincing that this paper is being a bit cowardly about the term multilevel selection, check out this line:
In other words, even though selection favors one species at the individual level (due to higher “invasion fitness”), it favors the other at the level of the overall community (due to community structure stability). I realize that when we think about multilevel selection we usually think about opposing fitness effects at the level of individual and group of a particular species, but perhaps this is too limited an interpretation of multilevel selection. In my eyes, multilevel selection is a valuable tool for thinking in general about conflicting pressures at different levels, and there is a lot to be learned by contrasting previously-posited mechanisms of multilevel selection with this new “stability selection”. I am not so sure that there’s anything all that novel here, but perhaps stability could be a better frame for these higher-level evolutionary processes.
Another place this paper falls short has to do with terminology. When you are trying to forge a new field of inquiry you are often charged with the unenviable task of choosing new terminology. So I am sympathetic to the difficulties here, but I am not overly fond with the choices made by these authors. The terms nonadaptive selection and systemic selection really do not describe the processes involved, and that does bode well for the ‘cultural fitness’ of these terms. Bad terms can hinder a fledgling field.
What’s my gripe? Well, first off these are not selective processes. This is a whole other topic that I cannot delve into in this post, but the influence of domestication (which actually is a selective process) on Darwin and Wallace’s choice of the term natural selection was lamentable enough. Nature does no selecting, either in sense of having intentionality or in the sense of advantaging traits in isolation. I would suggest that thinking about the stability of an individual in its environment could have been at least as productive a metaphor for describing what we now call adaptive selection for individual traits. But even if we accept adaptive selection for individual traits as a useful metaphor for what happens at this level, the value of the term ‘selection’ totally dissipates at levels above the individual. The authors of this review paper recognize this problem on the coarse scale, but they get the fine scale distinction totally wrong:
I disagree! After all, if the environment that atoms have to ‘survive’ in consists of subatomic forces that are inherent to the universe itself, I would absolutely be fine with saying that hydrogen is better adapted to the universe. What I would be more unwilling to say is that hydrogen was selected by the universe. The selection metaphor is tortured enough at the individual level: why import it — and all of its shortcomings — to describe an evolutionary process that is clearly not selective at all?
Another big issue here is how the environment is defined. This paper makes a potentially-interesting point imprecisely, which is that while individual adaptations evolve in response to environmental conditions, systems can evolve independent of their environment. I would be fine with this statement if it specified the physical (in other words, abiotic) environment. It is to some degree true that social systems and community structures may be stable or not stable independent of the physical environment in which they are found. But just saying “environment” ignores a major understanding of adaptive evolution, which is that other organisms make up a large part of the environment. So to say that a community structure can evolve independent of its environment is just silly: its potential structure, dictated by the presence or absence of different species, is completely a function of which species are in the environment, including potential immigrant species. I guess that you can think of the entire structure as a unit of selection and therefore not creating its own environment, but even this interpretation ignores the fact that communities may exert environmental influence on adjacent communities. And I am even worried that to exclude the physical environment might not make sense either: a particular community structure might be stable in a cold environment and unstable in a warm environment if those abiotic differences impacted the interaction strengths between species in each community.
I don’t know what to call the evolutionary process this review paper nicely explores. We need to work on terminology. In my eyes evolution of systems does involve the environment and does not involve selection. I will let more clever folks take it from there, but I don’t like the term systemic selection at all.
Finally, there is a really important implication of the idea this paper presents that is not represented at all in the paper. What these authors call adaptive selection is very likely to be strongly influenced (if not completely dictated) by the configuration of the stable system in which individual species and organisms can persist. In other words, the environment to which natural selection allows organisms to evolve is in large part the result of what systems are stable. Once an organism can survive in its physical environment, the rest is all about surviving in the environment of its interactions with both conspecifics and heterospecifics. These interactions are often indirect and yet can have a profound effect on whether an individual member of a species can survive. Broad translation: if the evolution of systems is really an important driver of the overall evolutionary process, the importance of individual selection and evolution of species in regards to which species persist and which species go extinct may be overblown. I am curious whether this implication of their major premise simply eluded these authors, or whether this idea was too radical to get the paper published in TREE. Given the smarts on this author team, my money is on the latter.
I was excited to learn from this review article that Lev and his long-time colleague John Damuth will be writing a book on so-called nonadaptive selection, as this is an evolutionary phenomena that needs to be treated more seriously.
I am also excited that this article was published. It represents a major step forward in thinking about how evolutionary processes actually shape ecosystems. I am glad to see such a large group of scientists — including many prominent researchers — support the idea that we need to get beyond simplistic evolutionary models that rely solely on the mechanism of selection on individuals. But I also see this paper as an initial hand hold in what still remains a very long and difficult climb. A part of that climb will involve harmonizing existing — but controversial — ideas about how different levels of selection operate to produce evolutionary change. We may also need to scrap old terminology — even the very idea of selection itself — in order to produce an evolutionary theory that explains not just the smaller-scale adaptations of individuals but the very structure of nature itself.
A Major Post, Adaptation, Articles, Community Ecology, Ecological Modeling, Ecosystem Ecology, Evolution, Evolutionary Modeling, Macroevolution, Multilevel Selection, Predation, System Stability