Recently we polled readers about their views on various controversial ideas in ecology. Here’s how I voted, with links to key references explaining my thought process. I share my thinking in the hopes of encouraging others to share their thinking as well. We’ll all learn something that way. And hopefully it’ll be fun (though I recognize that disagreeing with other people about science on the internet isn’t everyone’s cup of tea, which is fine.)
Recall that the poll asked respondents to indicate on a 5 point scale whether each idea was definitely false (1), definitely true (5), or somewhere in between. Respondents were asked not to provide an opinion on any idea about which they knew nothing, and to indicate their level of expertise on each idea (expert, know something, know a bit, know nothing). Ideas are listed in rough descending order of how controversial they turned out to be among poll respondents.
Species interactions typically are stronger and more specialized in the tropics: I know a bit, and picked 2 (probably false). Jeff Ollerton and Angela Moles’ guest post and subsequent paper and references therein make me dubious that this idea “typically” holds. Also, the discussion around this idea both on our blog and in the peer-reviewed literature (Anstett et al. 2016) suggests to me an idea that’s pretty vague. There seems to be significant disagreement as to exactly what’s being claimed and why we might expect it to be true, theoretically. Which leaves many “researcher degrees of freedom” to decide exactly how to operationalize and test the idea. I don’t think vague ideas pan out very often in ecology (or any other field, such as social psychology). They can prompt a lot valuable research, in part because vague but interesting-sounding ideas attract lots of people to work on them. But the outcomes of that research don’t tend to include the discovery of clear-cut, general empirical patterns, or well-developed theories that can be tightly linked to data. My “prior” is that any claim that’s so vague that experts disagree how to operationalize and test it will probably not turn out to be “typically” true. But I’m sure there are exceptions. And maybe my impression of the vagueness of this idea just reflects my comparative ignorance. Looking forward to your comments, as always.
Local biodiversity is declining in most or all localities: I’m an expert, and I picked 1 (definitely false). Vellend et al. 2013 and Dornelas et al. 2014 compiled the available data, which show that this idea is false. I am aware of the criticisms these papers have received. But as Vellend et al. (2016) show, those criticisms are based on a mixture of technical mistakes, and debatable arguments that seem unlikely to overturn the key results. I think the available data on this question are sufficient to constitute a severe test that this idea failed to pass. (Philosophical aside: Further data might of course change that conclusion–or might not! But that’s always true, about every question in science (and life). We could be wrong, and our conclusions might change in future. I don’t think that fact means that scientists should never, ever characterize any idea as “definitely” true or false. If asked “What’s the speed of light”, you should answer “186,000 miles per second”, not “Well, our current best estimate is 186,000 miles per second, but we could be wrong andfuture data might change that estimate.” Routine, ritualistic expressions of doubt about every statement don’t serve any purpose. So I respectfully disagree with any poll respondents who refused to pick 1 or 5 for any of the questions on the grounds that scientists should never say that any scientific claim is “definitely” true or false. But I’m not sure there were any such respondents, so maybe this aside is pointless.)
Species’ poleward geographic range limits typically are set by abiotic factors, not species interactions: I know something, and I picked 2 (probably false). Louthan et al. (2015) say that this idea “has not been clearly defined or broadly tested”. And see the excellent review of the limited evidence from transplant experiments in Hargreaves et al. (2014). Also, abiotic factors and species interactions probably aren’t best thought of as alternatives here, because they’ll interact in various ways. On their own those considerations would have led me to pick “2” or “3”. I went with “2” because the underlying logic of this hypothesis has never quite made sense to me. The basic idea, as I understand it (and perhaps I’m totally off-base here, since I’m not an expert!), is that “harsh” environments such as at northern latitudes, densities of competitors will be low. If competitor densities are low, interspecific competition won’t much reduce a species’ per-capita growth rate when rare. Therefore a species’ northern range limit will be found wherever it’s able to tolerate the local abiotic conditions just well enough to break even demographically. Leave aside the potential empirical problems with this, such as that an environment that’s harsh for one species might be benign for its competitors, or the potential role of dispersal in setting range limits. This idea doesn’t work even on its own terms. In a harsh environment, a species’ per-capita growth rate when rare will be low even in the absence of interspecific competition (that’s presumably what “harsh” means). Which means that interspecific competition doesn’t need to much reduce the per-capita growth rate when rare to push it into negative territory. So even if interspecific competition doesn’t much reduce your per-capita growth rate when rare in harsh environments, well, it only needs to reduce your per-capita growth rate when rare by a little bit for you to get competitively excluded. The intuitively-appealing but logically-flawed idea that anything that tends to hold down species’ densities–“harsh” environments, high density-independent mortality rates, frequent and/or intense disturbances–thereby prevents competitive exclusion is one of the common mistaken ideas in ecology (Chesson and Huntly 1997 (great paper), Violle et al. 2010, Fox 2012, and see here).
The intermediate disturbance hypothesis (empirical version): species diversity typically peaks at intermediate disturbance frequency or intensity: I’m an expert, and I picked 1 (definitely false). Mackey and Currie (2001) reviewed the empirical literature on the IDH. According to their results this idea is false; peak diversity at intermediate disturbance isn’t “typical”. Even controlled experimental manipulations of disturbance often fail to find a humped diversity-disturbance relationship (reviewed in Fox 2012). The meta-analysis of Svensson et al. (2012) does find that, when you measure diversity as species richness and fit a quadratic regression to the diversity-disturbance relationship, most published studies estimate a negative quadratic term. Trouble is, quadratic regression has recently been shown to be a biased test that almost certainly generates many false positive humps. Concave-down bivariate relationships tend to be misdiagnosed as humped by quadratic regression.
The intermediate disturbance hypothesis (one theoretical version): disturbances promote diversity by interrupting competitive exclusion and preventing the community from achieving equilibrium: I’m an expert, and I picked 1 (definitely false). See Fox (2012) and Chesson and Huntly (1997) for explanation of why this idea is definitely false. Note that there are of course “fluctuation-dependent” coexistence mechanisms by which periodic disturbances can generate stable coexistence that would not occur in the absence of those disturbances (see the series of posts starting here). Perhaps fluctuation-dependent coexistence mechanisms were what some poll respondents had in mind when they characterized this idea as probably or definitely true? In my view, summarizing fluctuation-dependent coexistence mechanisms by saying that disturbances or other environmental fluctuations “interrupt” competitive exclusion is a bad way to summarize them.
Biodiversity is among the most important determinants of ecosystem function: I know a bit, and I picked 3. I confess I was lazy on this one. I only vaguely recalled the relevant review papers and couldn’t be bothered to look them up, so I picked “3”. I’ve since gone back and looked up Hooper et al. (2012), Tilman et al. (2012), and Duffy et al. (2017). On that basis, I’d now say I know something and might lean towards “4” (probably true).
Habitat fragmentation per se (as distinct from habitat loss) typically reduces biodiversity: I know a bit, and I picked 2 (probably false). And honestly, you could probably talk me into “1” (definitely false). Fahrig (2017) reviewed the relevant literature and found that 3/4 of significant effects of habitat fragmentation per se on biodiversity are positive. Which you wouldn’t know if you only read the abstracts of published studies. Fahrig (2017) found that authors writing on this topic systematically skew their findings in their abstracts, often not bothering to mention positive effects of habitat fragmentation on biodiversity even when the only significant effects they found were positive.
Interspecific competition can be reliably inferred from information about functional trait similarity: I’m an expert, and I picked 1 (definitely false). Yes, I know that there are famous cases like Darwin’s finches, in which there’s a clear, interpretable mechanistic link between key phenotypic traits used in feeding, and the occurrence of interspecific competition (in Darwin’s finches, the key traits are beak size and shape). Threespine stickleback in postglacial lakes in British Columbia are another classic example. And there are others. But such examples are rare. Most attempts to infer interspecific competition from “functional traits”, or even to use functional traits to explain the outcome of interspecific competition that’s been demonstrated via other means, fail. For instance, Kraft et al. (2015). I speculate that that’s in part because many “functional traits” get measured because they’re convenient to measure, not because they have a tight mechanistic connection with competitive outcomes.
Species diversity typically peaks at intermediate productivity: I’m an expert, and I picked 1 (definitely false). My reasoning here is more or less the same as for the empirical version of the IDH. Not enough empirical studies find humped diversity-productivity relationships in enough systems for a humped relationship to be considered “typical” (Mittelbach et al. 2001), and many putatively-humped relationships probably are false positives because quadratic regression is a biased test for humps.
Species richness on continents is dominated by ecological limits rather than evolutionary limits: I know something, and I picked 3 (not sure). I learned about this controversy from close reading of the papers arising from the Oxford-style debate at the ASN meeting a few years ago. I thought that debate was very interesting. I partially agreed and partially disagreed with both sides, while also thinking the question might not be sufficiently well-posed to be resolvable.
Higher biodiversity will buffer ecosystem function against climate change: I know a bit, and I picked 3 (not sure). This is another one on which I was lazy and didn’t bother to remind myself of what the relevant studies actually found.
Predator functional responses typically are ratio-dependent: I’m an expert, and I picked 1 (definitely false). First of all, it’s important to remember that ratio-dependence is a very special case of functional responses that depend on both predator and prey densities. Published measurements or estimates of functional response shape often find that per-predator, per-prey feeding rate depends on both predator and prey density in some fashion, but they rarely (never mind “typically”) find ratio dependence (DeLong and Vasseur 2011, Schenk et al. 2005, Zimmerman et al. 2014, Nowak et al. 2017). Second, other lines of empirical evidence for ratio-dependent functional responses are very weak and indirect. The fact that ratio-dependent predator-prey models can be fit to (e.g.) data on changes in trophic level biomasses along productivity gradients is not evidence for ratio-dependent functional responses, because all sorts of other mechanisms having nothing to do with functional response shape can explain those data. Third, the theoretical arguments for ratio-dependent models strike me as…idiosyncratic, and I agree with Peter Abrams’ criticisms of them.
Top-down effects typically are stronger and more important than bottom-up effects: I’m an expert, and I picked 3 (not sure). Basically, I picked “3” for reasons 5 and 6 on this list. I’m not sure this question actually makes sense as stated, though it could be modified to make sense. And on theoretical grounds, we should expect the strength of top-down effects to depend on basal resource enrichment, and the strength of bottom-up effects to depend on food web structure (e.g., Oksanen et al. 1981, Abrams 1993, Leibold 1996).
Species’ equatorward geographic range limits typically are set by species interactions, not abiotic factors: I know something, and I picked 4 (probably true).
My reasoning on this one is analogous to my reasoning for poleward geographic range limits, discussed above.
The dilution effect: host biodiversity generally reduces disease prevalence or infection risk: I know a bit, and I picked 2 (probably false). My impression from the occasional papers I’ve read, occasional comments from others on this blog, and chatting with an expert (our own Meghan Duffy) is that the dilution effect occurs in some systems but not others. So based on what little I know, I think it’s probably false to say that host biodiversity “generally” reduces disease prevalence or infection risk.
Ecological networks (e.g., food webs, plant-pollinator networks) have the structures they do because other structures would be unstable: I’m an expert, and I picked 1 (definitely false). From reading the literature, the experience of a former student of mine trying to document an alpine plant-pollinator network (Olito and Fox 2014), and speaking with experts who are using new methods to better sample “who pollinates whom”, I am dubious of the empirical evidence base on this. That’s not a criticism of the many outstanding empirical researchers who work on this; it’s just a really hard thing to sample. I suspect published plant-pollinator interaction networks badly undersample rare interactions, and I’m sure I’m not alone in that suspicion. I also think that pollinator choices about which plants to pollinate often are quite variable and flexible in space and time, at many scales (the same could be said of many predators and their choice of prey). So I worry that theoretical models trying to explain observed plant-pollinator network structures on the grounds that other structures would be unstable are in a similar position to food web theory back in the late ’70s and ’80s, when the available comparative data on food web structure weren’t good enough to theorize usefully about. I also think that persistence of plants and their pollinators, and predators and their prey, depends on various factors besides interaction network structure. Just because network structure A is less likely to be stable than network structure B, all else being equal, in a theoretical model in which other persistence-affecting mechanisms are ignored, doesn’t necessarily mean that structure A is unlikely to be observed, or that structure A is less likely to be observed than structure B (see here for related discussion). Finally, theory provides mixed predictions on whether stable interaction networks have more realistic (or even different) structures than unstable ones (e.g., Kristensen 2008). The answer seems to depend on details of model assumptions, including but not limited the chosen measure of stability.
The enemy release hypothesis: invasive species often spread and become abundant relative to natives because enemies attack the invasives less: I know nothing and so expressed no opinion.
Hubbell’s neutral model: community dynamics typically are dominated by drift, because all individuals of all species typically are demographically equivalent: I’m an expert, and I picked 1 (definitely false). As I think is widely recognized at this point, the fact that neutral models can fit empirical species-abundance distributions is no evidence for communities actually being neutrally stable, so that drift dominates community dynamics. The shape of the species-abundance distribution is not informative about the processes that generated it (e.g., McGill 2003, and see this old post). And there are of course all sorts of other lines of evidence showing as conclusively as anything can ever be shown in ecology that community dynamics are not “typically” dominated by drift (although drift does sometimes matter; Gilbert and Levine 2017). Just off the top of my head (because if I tried to be comprehensive we’d be here all week): Yenni et al. (2017), Usinowicz et al. (2017), Kraft et al. (2015), Harpole and Tilman (2006), Siepielski et al. (2010), Silvertown et al. (2006), the various lines of evidence reviewed in Siepielski and McPeek (2010), Hille Ris Lambers et al. (2012) and Vellend (2016)…
Interspecific competition and facilitation can be reliably inferred from co-occurrence matrices or correlation matrices of species’ abundances: I’m an expert, and I picked 1 (definitely false). See here for my reasoning. And see Sander et al. (2017), Barner et al. (2018) and Freilich et al. (2018), finding that a wide range of methods based on co-occurrence or correlational data completely fail to infer species interactions that are known to occur in a well-studied natural system (the rocky intertidal). Any method for inferring species interactions from observational data that fails when applied to data from the rocky intertidal–a system famous for its intense species interactions!–is unreliable.
Interspecific competition can be reliably inferred from information about phylogenetic relatedness: I’m an expert, and I picked 1 (definitely false). See Mayfield and Levine (2010) and other lines of evidence, briefly reviewed here.
Population growth typically is density-dependent: I’m an expert, and I picked 5 (definitely true). Time series analyses almost invariably estimate density dependence, though it’s often often too weak for the 95% c.i. to exclude zero, given the typical length of ecological time series and typical levels of sampling error (Ziebarth et al. 2010, Knape and deValpine 2012). The preponderance of experiments that have looked for intraspecific competition find it. And if population growth weren’t typically density-dependent, species typically would either increase to impossibly high abundances or else go extinct, in much less time than the typical species lifetime as estimated from the fossil record (millions of years).
Earth is currently experiencing a 6th mass extinction: I know a bit, and I picked 2 (probably false). Paleontologist Doug Erwin studies mass extinctions, and his arguments against the claim that we’re in the midst of a 6th mass extinction seem reasonable to me.