One of the most fraught questions in ecology is whether or how ecologists should seek generalities. Debates over this broad question crop up in many contexts and take on many forms. Think of debates over whether ecology has “laws” analogous to the laws of physics. Debates over whether ecologists ought to focus on producing information relevant to management or conservation of specific species or locations, rather than on less-useful generalizations. Debates over whether we should give up on doing community ecology because every community is an idiosyncratic special case. Etc.
In an old post, I tried to partially resolve some of these debates by suggesting that there are many different sorts of “generality” that ecologists might seek. I argued that they’re all valuable in their own way, though some might be more achievable and/or valuable than others depending on the goals and interests of the investigator.
But it’s my anecdotal impression that ecologists as a group value some forms of generality over others. And that different sorts of ecologists tend to seek, and value, different sorts of generality. To get some data, I hope you’ll take the short poll below (just four questions). It asks you about your opinions about, and own use of, various sorts of “generality” in ecology and ways of seeking generality. Here’s a brief summary of each of them:
- Universal or nearly-universal patterns or “laws”. Think of quarter-power body size allometries, the latitudinal species richness gradient, the species-area curve, Bergmann’s Rule, etc. They provide generality because many different systems/species/cases fit the same pattern, or obey the same rule or “law”. (EDIT: as Brian points out in the comments, Bergmann’s Rule turns out to be more of a purported pattern or law than an actual one. It has too many exceptions to be a pattern or law. But I’m leaving it here as an example, because sometimes when you set out to study a purported universal pattern or law, you find that it’s not a pattern or law at all. But you still set out to study a (purported) pattern or law, so your research still falls in this category.)
- Meta-analysis. That is, statistical summaries of the results of different studies of the same phenomenon. They provide generality in the sense that they tell us what’s statistically “typical” or common, and how much variation there is around what’s “typical”.
- Simple theoretical models. I’m thinking here of theoretical models that are intended to “capture the essence” of the phenomenon being modeled, “sharpen our intuitions” about the phenomenon being modeled, or identify some general “principle” about the phenomenon being modeled. These models aren’t intended as realistic, exact descriptions of any particular system, in part because they assume away or make very simple assumptions about other phenomena besides the one being modeled. But these simple models are thought to provide generality because they apply in an approximate way to many different systems, or act as a simplified “limiting case” for many different systems. Think of the Lotka-Volterra competition model, the Rosenzweig-MacArthur predator-prey model, Tilman’s R* model of resource competition, the marginal value theorem of optimal foraging, etc.
- Statistical attractors. Think of MaxEnt-type phenomena: empirical patterns that are common because they’re hard to avoid. The pattern is commonly observed (and so “general” in that sense) because many different ecological processes/mechanisms/scenarios would give rise to the pattern, and few would give rise to any other pattern. The pattern is thus a statistical “attractor”–a statistical inevitability. For instance, the fact that species-abundance distributions always have a lognormal(ish) shape may indicate that that shape is a statistical attractor. That shape is common because population growth is always a multiplicative stochastic process, and system-specific ecological details have little or no effect on the overall lognormal(ish) shape of the distribution.
- “High level” theoretical frameworks. Think of modern coexistence theory, the Price equation, and the “four fundamental forces” framework of population genetics (selection, drift, mutation, migration). These “high level” frameworks provide generality in the sense that they unify and subsume various models as special cases.
- Fruitful analogies. I think of this as the Tony Ives approach to generality. For instance, lots of different ecological systems have been hypothesized to exhibit alternate states, with stochastic perturbations occasionally flipping the system from one state to the other. In the linked post, Tony Ives talks about how his previous experience with alternate states in other systems helped him recognize and model the possibility of alternate states in the population dynamics of Icelandic midges. That is, there are analogies one can draw between those midges, and other systems (including even non-living systems!) that exhibit alternate stable states. Those analogies are a form of generalization–they allow us to reinterpret our knowledge of one system so that it applies to another, analogous system. As another example, think of how Steve Hubbell’s neutral theory, and Mark Vellend’s theory of ecological communities, both start by drawing an analogy between community ecology and evolutionary biology.
- Model systems. A model system is one that has features that make it particularly tractable to address the question of interest. Studies in model systems often are thought, hoped, or assumed to apply to other systems as well. Sometimes because the question of interest isn’t tractable to address in any non-model system.
- Distributed experiments. That is, experiments that run simultaneously at many sites, using the same methods everywhere, so as to facilitate cross-site generalization. One could consider this a special case of meta-analysis, but I split it out because I feel like many ecologists tend to think of it as its own thing. Think of NutNet, for example.
- Long-term studies. A long-term study should capture a greater range of temporal variation than a short-term study, and so should provide insights and conclusions that can be applied in a greater range of circumstances.
- Other. No doubt my list of forms of generality and ways of seeking generality is incomplete, at least in the eyes of some of you!
Looking forward to your responses!
Dynamic Ecology 
Interested to see the outcomes. I bet I’m in a minority in favoring statistical attractors and nearly universal laws over some more traditional approaches.
But I’ve always appreciated this list you made.
FWIW I would not lump Bergmann’s rule in with SARs and 3/4 scaling. With growing data, Bergmann’s rule seems to be a 50/50 proposition which is not too impressive given there are only two outcomes.
yeah, I bet that’ll place you in the minority too.
I’m also especially curious about unifying theoretical frameworks, which I personally love. I find it hard to guess how popular they’ll turn out to be.
Yes, I should maybe clarify that some of those universal patterns/laws are actually *purported* universal patterns/laws. Sometimes when you study the purported universal pattern/law, you discover it’s not universal after all.
I bet “working in model systems” is actually going to turn out to be the least popular way to seek generality. Both in terms of how many people use it themselves, and how important they think it is for ecologists as a group to do it.
Of course we have to seek generalities, not to make ecology just to be a stamp collection. However, one may question how fruitful it is to strive after laws in the same way as done in physics. I prefer to view them as commonalities that help me understand any type of system that I am looking into. In the same way, I prefer to view ecological theories as a toolbox that I can use for interpreting patterns and process within any particular system. Sometimes I need a hammer and other times a screw driver, but if I lack any of these items in the toolbox then I may be less able to reach an understanding. So, research to me is to build on this toolbox through the understanding of general patterns whether these are from empirical data or from simple population models AND to use this toolbox to understand any particular system/problem. With that said, some ways of seeking generality, as you describe, are more fruitful than other. In this respect, I generally find the first half of the list to be more effective for general understanding than the second half but that may of course also depend on your favourite type of question.
“Of course we have to seek generalities, not to make ecology just to be a stamp collection. ”
There are those who’d disagree! Think Shrader-Frechette & McCoy’s book Method in Ecology, for instance. Or this Peter Kareiva essay, in which he argues that ecology is (or should be) about applying universally-applicable *methods* to unique particular situations: https://www.nceas.ucsb.edu/nceas-web/projects/resources/ecoessay/brown/kareiva.html. That is, ecology’s a stamp collection, but stamp collectors can use universally-applicable methods to identify any stamp, figure out what it’s worth, etc.
Kareiva’s not too far from your “toolbox” view, I don’t think. Except that he doesn’t think it’s all that important build the toolbox by trying to understand general patterns. He thinks we can build the toolbox just fine in other ways.
Would a theoretical framework count as a universally applicable method? Eg things in my ‘toolbox’ are density dependence, Allee effects, population structure…
Doing more theory-data integration, I predominantly expand theoretical frameworks to either (a) interface them with long-term data (with a few relevant system-specific tweaks layered on), or for more theoretical work (b) inject bits of biological realism as resolved in meta-analyses, laws, or earlier analogies. As I like putting things into boxes, I think all of these approaches are crucial and biased towards frameworks.
Random: many studies empirically resolving a process end up happening in more-or-less model systems. Also not sure if I’ve used any statistical attractors beyond the fact that “process error” tends to be multiplicative and lognormal.
Yes, good question where to fit a broadly-applicable abstract concept like “density dependence”, “Allee effect”, or “population structure” in this little poll. One could argue that they’re their own thing and I should’ve included a separate category for them. I might argue that myself, now that you’ve pointed it out to me! 🙂 Or you could put them under “simple theoretical models” and/or “high level theoretical frameworks” depending on exactly how you make use of them.
@ Vadim:
Thinking about this further, I might say that abstract, broadly-applicable concepts like “density dependence”, “population structure”, etc. are a prerequisite for any of the forms of, or approaches to, generality on my list. For instance, you can’t conduct a meta-analysis unless you have some variables that you can measure in many different studies. Those variables have to be defined in a sufficiently abstract way that they’re measurable on different things. If you’re doing a distributed experiment that involves herbivore removal, well, that experiment depends on there being an abstract category of things called “herbivores”. If you’re studying the universal law that resting metabolic rate scales with body size raised to the 3/4 power, well, you’re using the abstract concepts of “resting metabolic rate” and “body size”. If you want to study the species-abundance distribution as a statistical attractor, well, that means there have to be many different entities that can all be called “species”, each of which has a property called “abundance”. Etc.
Does that make sense? (I feel like I’m not being very clear…)
Sorry, I was a bit sloppy with my examples – I wanted to point out fundamental processes (density dependence, Allee effects) as a potentially distinct category that I interpreted as ‘simple models’. Population structure is just a biological feature (like the other variables you mentioned) that can strongly mediate the relevance/effect of specific processes
I’m not sure if having ‘fundamental processes’ in place of ‘simple models’ is logically sound, but I think this emphasis reflects an important path to generality in ecology. Pure theory highlights when mixing different processes and features together can yield surprising patterns/dynamics. On the empirical side you might quantify the role of each process using either experiments or (with important caveats) compare the fit of models combining different processes. This (1) gives system-specific insights and (2) resolves the conditions under which each process can have strong effects across species/systems/environments – which is critical for generality. Unfortunately, the second benefit is very often lost during study design or publication as most funding opportunities and applied research remain framed around species/systems.
Also – I realized I accidentally lumped simple models with high-level frameworks in my mind. I guess frameworks are very satisfying and pedagogically helpful but I’m not familiar enough with them to say whether they add useful structure to theoretical ideas, a bit like how delineating communities is often of dubious benefit.
Well, I suppose that depends on what you mean by a generalisation. I guess that Kareiva would not disagree when saying that theories are important for understanding ecological patterns, but theories are also generalisations. But, I am not sure what you mean when you talk about universally-applicable methods as an opposition to generalisation.
I think that what you describe as stamp collection for ecologist can be viewed as some kind of problem solving without trying to understand underlying mechanisms. That is fine with me, and probably needed at times, but it seems to me like something else than ecology as a science. And, I do not think that is what Kareiva is arguing for.
Hi Jeremy; I have always appreciated your list of ‘seeking generality in ecology’. 3 points.
First,Coming from a fisheries background, I appreciate the Kareiva perspective; there are many specific applied questions that call out for specific answers. And we probably have the tools to put together useful answers….given commitment to gathering the right data. But we often cant/dont do this. ON THE OTHER HAND, a great many applied questions in fisheries make use of generalities that were found via the other methods you list. Or these other rules are embedded in bigger fisheries models;Scaling rules for many things underlie many fisheries simulation models. Metabolic ecology rules for temperature, as well as size, are widely used , and so forth. these are widely used because we don’t have system specific data ; much literature on ‘data poor fisheries’ decision making now exists.
Second; your list mixes 4 things: how do we find rules? what kind of rules do we look for? How do we test rules? and Do we dig for deeper explanations for the rules we find[ and how]? Sometimes its useful to mentally separate these 4. Example: There are several big useful scaling rules in fisheries validated via comparative studies [ of many kinds; space, time, across species, etc] that are useful despite lacking any deeper explanation.
third, My own research uses all of the above [ except distributed experiments] In various cases. But to always be looking for a generality: and then to look for deeper PRINCIPLES that can then inform even more situations is the game of choice for me.[ of course see the Marquet,et al 2014 ON THEORY IN ECOLOGY paper{https://academic.oup.com/bioscience/article/64/8/701/2754269}. ]
In the 1970s-80s there was a revolution in thinking about plant evolutionary ecology; caused by careful thinking with ESS models of individual selection….often using at its core the sex allocation principle of ….everyone has one mother and one father[ due to RA Fisher]. This was blended with much older ideas [ due to Darwin and RA Fisher] about the core problem plants faced of avoidance[ or not] of selfing [ another principle] to produce a very powerful new approach. how simple, how powerful. Principles played a big role in this.
The Marquet link https://academic.oup.com/bioscience/article/64/8/701/2754269
“a great many applied questions in fisheries make use of generalities that were found via the other methods you list… these are widely used because we don’t have system specific data”
Yes. I think that’s an oft-overlooked reason for applied ecologists to care about general empirical patterns, and about simple theoretical models. We need *something* to fall back on in the absence of system-specific information. Another example is that, in many jurisdictions, rules for evaluating the extinction risk a species faces often are based in part on “rules of thumb” derived from simple stochastic population models from Russ Lande and others.
“Second; your list mixes 4 things”
Yup. I think a lot of interest centers on how to mix & match those things. For instance, what I think of as the “MacArthurian” approach to ecology starts by trying to identify an empirical pattern or “law”, then proposes a simple theoretical model to explain it. The generality of the pattern–the fact that it applies in many different cases or systems–suggests that its explanation is independent of system-specific details, hence motivating a simple theoretical explanation that assumes away system-specific details. But the “failure mode” of that approach–the most common way it goes wrong–is when the pattern in question is a statistical attractor. If you identify a statistical attractor, and then propose a simple theoretical model to explain it, you’re in trouble. A statistical attractor is often observed not because there’s one underlying mechanism or process that generates it, but because pretty much any underlying mechanism or process would generate it. Think of the failed attempt to explain the universal lognormal-ish shape of the species-abundance distribution by appeal to a very specific simple model, neutral theory.
I find it really interesting that on twitter everybody is acting like its a totally settled question to pursue generality. While I certainly agree, that is not my read of the state of the field. MacArthur admonished his peers of the need for generality in his death bed book in 1972. And we still have national academy members writing society presidential addresses like this: https://www.jstor.org/stable/10.1086/420777?seq=1#page_scan_tab_contents which is basically an attack on the idea of general principles in science – both their existence and their relevance. And then there is just the body of what is getting published, most of which is not especially general.
Of course its important to note that while applied science benefits from general results, there is also a need for non-general results (if I’m going to manage a deer population I need to know how big it is and what the main drivers of its population size are – and that is very space/time/taxon specific). I think this is probably the root of both Kareiva and Simberloff’s thinking.
While those who don’t click through and read your list of ways to get to generalities and the poll are missing something, I hardly think its as a trivial question to ask the more basic question of the place of generality in ecology both under current practice and in a potential ideal mix as people are making it out to be.
I hadn’t seen the Twitter commentary on this post and don’t have time to look right now. But I can say a few things that hopefully will advance the conversation (or maybe I’ll just be repeating things others have already said elsewhere; repetition is an inevitable byproduct of a distributed conversation…)
Based on the poll results so far, which admittedly are far from a random sample of ecologists, ecologists who don’t seek any sort of generality in their own work are very rare. So whether to pursue generality *in some sense or other* does seem like a settled question. Though if that’s right, I do wonder how long it’s been settled. We don’t have comparable poll data from 10 or 20 or however many years ago. I’m certainly open to the possibility that there are a few prominent voices in the field (e.g., Dan Simberloff, Peter Kareiva, or further back, Schrader-Frechette & McCoy) whose views on “generality in ecology” are unrepresentative and uninfluential.
I think the more interesting issue, and the one this poll is trying to get at, is that “generality” is a broad term that means different things to different ecologists. As the poll results will show, there certainly is a lot of variation among ecologists as to what sorts of generality they pursue in their own work. And a lot of disagreement among ecologists as to what sorts of generality are most important for ecologists collectively to pursue. There’s also anecdata for this. For instance, when Tony Ives asked for a show of hands at his MacArthur Award lecture a few years ago as to how many audience members thought ecologists should seek to discovery general laws, only about 1/3 of the hands went up. That was an audience of several hundred ecologists at the ESA annual meeting. So while not a random sample of all ecologists it’s definitely a large and diverse enough sample that it shouldn’t be dismissed lightly.
I also think there’s a tendency in ecology for people who seek *one particular sort of generality* to write off others as uninterested in generality at all, rather than being interested in *some other sort* of generality. Or maybe not–maybe that too is something that only a few prominent but unrepresentative ecologists tend to do? As I noted elsewhere (https://dynamicecology.wordpress.com/2019/09/05/ask-us-anything-good-vs-bad-provocation/), provocative claims that lots of other ecologists are Doing Ecology Wrong (e.g., by reveling in idiosyncratic details and not seeking generality) have tended to come from a few prominent white guys who like to argue. So if you only look at dueling opinion pieces about how ecologists aren’t seeking generality, or aren’t seeking the “right” kind of generality, maybe you get the mistaken impression that there’s much more disagreement on those questions than there actually is. Because you’re only reading the views of argumentative people who hold extreme views.
The *really* hard questions are (i) whether ecologists collectively pursue the “ideal” mix of different forms of generality, and (ii) whether there are some forms of generality that ecologists need to get better at pursuing. I don’t know the answers to those questions, though of course I have my own opinions.
Pingback: A brief history of ecologists’ disagreements about “generality”, in quotes | Dynamic Ecology
Pingback: Poll results: the many ways ecologists seek generality (and why some are much more popular than others) | Dynamic Ecology