Trying to save a zombie idea (UPDATED)

In a previous post I commented that it would be interesting to see whether the very nice paper by Adler et al. (2011) on diversity-productivity relationships in terrestrial grasslands would finally kill off the zombie idea that diversity generally is a humped function of productivity. I also suggested that this zombie might be more of an apparent zombie than a real one, because hardly anyone these days still believes in this zombie idea, it’s just that everyone thinks everyone else believes in it.

I was wrong on both counts (you’d think I of all people wouldn’t be so quick to see a zombie idea as dead or dying!) Writing in Science this week, Pan et al. and Fridley et al. criticize Adler et al., and the reply is here. Pan et al. claim that, properly analyzed, the Adler et al. data show a strongly linear diversity-productivity relationship, while Fridley et al. claim both that the data are “clearly deficient” as a test for a humped diversity-productivity relationship, and that they show a “clear” humped diversity-productivity relationship*.

In my view, the comments are a striking illustration of just how far even very good ecologists will go in an attempt to save a pet hypothesis in which they are heavily invested. I should emphasize in saying that that I don’t have a dog in this fight. The question of how plant diversity varies as a function of primary productivity is a purely empirical question. Addressing it properly requires careful study design to sample the full range of natural variation, control for confounding factors, etc. And personally I don’t care what the answer is. I just care that we get the right answer.

I think Adler et al. is the right answer, for grasslands. Pan et al. don’t, but as Adler et al. point out their comment basically amounts to cherry-picking data to obtain a linear relationship. Not consciously cherry-picking of course. But as a rule, in any large and complex dataset there is always some subset of data showing a relationship between variables different than that shown by the dataset as a whole, and you can always find some more-or-less-plausible post-hoc reason to pay special attention to that subset. As Grace et al. point out in their reply to Pan et al., Fridley et al. actually prefer to focus on a different subset of the data. That strongly suggests that both sets of commenters are, unconsciously, just looking for reasons to make the data show what they think the data ought to show.

As noted above, the Fridley et al. comment is, on it’s face, self-contradictory*. Fridley et al. also try to change the question, arguing that “productivity”, or any index of it, should be defined to include leaf litter, which would be contrary to the bulk of previous empirical and theoretical studies. They also argue that Adler et al. should have conducted various alternative analyses…that Adler et al. actually did conduct and presented in their paper. And finally, as Grace et al. show, what little “humpiness” there is in the Adler et al. data reflects the log-normal distribution of the data, so it looks slightly humped on an arithmetic scale.

I’m not an expert on this stuff (I’m not a plant guy), but I’m confident I know enough about it to come to an informed (not infallible, but informed) opinion. And in my view, the Adler et al. paper is totally unscathed. In grasslands, I think we now have the best data we can ever expect on the diversity-productivity relationship–collected at a large number of sites around the world, with good coverage of a big productivity range, sampled using a consistent protocol. And those data have been very carefully and thoroughly analyzed. The answer is clear: the diversity-productivity relationship is very weak and noisy, and it’s not humped. So we’d better learn to deal with it.

Even if you want to say that the humped diversity-productivity relationship is so well-established that we need extraordinary evidence to reject it, well, I disagree with your premise, plus Adler et al. is extraordinary evidence.

At the end of their comment, Fridley et al. conclude with some quite striking rhetoric. They claim that the hump-backed diversity-productivity relationship is a “cornerstone of plant ecology”, backed by “decades of careful mechanistic analysis”, and is “used by plant conservationists and restoration ecologists, as well as theoretical ecologists.” The first claim must surely be loose language on the part of Fridley et al., because taken at face value it’s patently false. Discarding the humped diversity-productivity hypothesis would not cause plant ecology to collapse the way a building would if you removed its cornerstone. If you think otherwise, your view of what “plant ecology” consists of is way too narrow. The vast bulk of research on plant physiological, population, and community ecology has nothing to do with the diversity-productivity relationship, humped or otherwise. I’m not sure what the second claim means, but I hope they’re not claiming that we’ve proven mechanistically that the diversity-productivity relationship has to be humped. There’s nothing in ecology like, say, the kinetic theory of gases that let’s us rigorously derive a diversity-productivity equivalent of the ideal gas law. And their other claims are irrelevant. If conservationists, restorationists, and theoreticians based their work on the assumption that the earth was flat, that wouldn’t make the earth flat (and yes, it is possible for many experts to be wrong about a matter of empirical fact). Worse, do you really want to argue that conservation and restoration practices, or theoretical modelers, should never adjust what they do in light of new data? Please don’t misunderstand me: the authors of the Fridley et al. comment include many very good, very smart ecologists (and Jason Fridley for one is a friend of this blog). I just find it striking that so many very good, very smart ecologists could be so reluctant to reconsider their position on what ought to be a straightforward empirical matter, and so willing to resort to what looks like quite poorly-supported rhetoric. Basically their rhetoric at the end amounts to saying “this zombie idea has been around for a long time–so therefore it can’t possibly be a zombie”.

UPDATE: Commentary from one of the co-authors of Adler et al. here.

*Grace et al. also note that Fridley et al.’s comment is self-contradictory. I agree with that characterization.