From the archives: what scientific claims have you changed your mind about?

New posts will be rare over the holiday period, so to tide you over I thought I’d highlight some old posts (going back to my pre-Dynamic Ecology days over on Oikos Blog) that didn’t attract much notice at the time. Starting with this one, in which I posed the question: what’s the biggest scientific claim about which you’ve changed your mind?

In the original post, I avoided answering my own question, so I’ll answer it now. In grad school, I was a pretty mechanistic guy, or at least I thought I was. Heavily influenced by having read some of David Tilman’s R* work as an undergrad, I thought that phenomenological models of any sort were just an expression of ignorance. At best, they were a sadly-necessary stepping-stone on the way to true (i.e. mechanistic) understanding of what was really going on. Nowadays, I recognize that any parameter or function in any model can be viewed as a “high level” phenomenological summary of some lower-level mechanism. And I appreciate that much of our most powerful and successful general theory (e.g., the theory of evolution by natural selection; modern coexistence theory) is phenomenological in precisely this way.

I avoided giving this answer originally, on the grounds that changing one’s mind because you were relatively-ignorant and then learned something doesn’t really count as changing one’s mind. But I’ve since changed my mind about that.🙂 While I certainly hadn’t read nearly as much as a grad student as I have now, I wasn’t totally ignorant either. I was capable of developing and defending a view of the role of modeling in ecology. Plus, it’s not as if I’ve now read everything, so that I’m now perfectly informed! So if I now have a different view on the role and value of different sorts of models in ecology, well, that means I’ve changed my mind.

6 thoughts on “From the archives: what scientific claims have you changed your mind about?

    • To clarify, I’ve always respected generalities. If anything, my growing appreciation for phenomenological models has increased my respect for generalities. Think of the Price equation, for instance, which is the most general mathematical expression of evolution by natural selection. It’s famously abstract and phenomenological–that’s precisely what *makes* it general.

      • Jeremy, I very much appreciate you posting this realization. I find many theoretically-oriented folks still have not arrived at the resolution you have described when you say “Nowadays, I recognize that any parameter or function in any model can be viewed as a “high level” phenomenological summary of some lower-level mechanism.” I believe this recognition is KEY in understanding the relationship between data and theory.

      • Happy New Year Jeremy, Brian and Meg!

        I just wanted to follow up on my last comment by saying that even though a parameter can be viewed as a “high level phenomenological summary of some lower-level mechanism” (Jeremy’s nice phrasing), it is possible to test ideas about the mechanisms that live behind that phenomenological summary. In a recent paper, Claire de Mazencourt and collaborators tested ideas about the mechanisms behind an observed effect of diversity on stability. I think you might find it interesting since it successfully implicated a number of lower-level mechanisms as collectively producing the high-level effect.

  1. Believe it or not, when I started out grad school (I was doing an MS) I was very skeptical of the continuum concept to explain plant species distributions. It took me a while to fully embrace it.

    Strange, but not so strange if you’ve been raised with Braun-Blanquet and phytosociology during your undergrad!

  2. This one really made me think. I suppose one could answer this question with two different interpretations of it in mind.

    With reference to scientific philosophies/methods, I would say a big one for me has been realizing the importance of not overfitting a model. I guess I used to vaguely assume that yeah, the more explanatory variables, and the higher the R^2, the better. Of course dummy, what else, you prefer to explain less of the total variance do you? Well, not necessarily the case. Similar to Barry Rountree’s original answer to the question I guess.

    A second one is a greatly increased appreciation of the importance of simulation analysis in obtaining insights of various kinds, especially for testing the zillion and one possible analytical methods people throw out there, but also of course for explaining the likelihood of observations under various possible models. I didn’t have a strong pre-existing opinion on the matter though, just didn’t really understand why people bothered with simulation when they could be out measuring stuff (“for God’s sake man, find something to measure and measure it!”)🙂

    With reference to actual subject matter topics, I used to trust the dendroclimatological estimates for the last ~millennium, but now I decidedly do not, because the assumptions and methods used to derive them are insufficient to the stated task, especially over long time scales (and in multiple, important ways). In the process, I’ve tentatively concluded that some (primarily physical) scientists simply don’t have a proper respect for the inferential problems stemming from complex systems–something that we biologists take as a fundamental tenet at virtually any hierarchical level of organization–and this lack can, and does, lead to very major errors of inference in something like paleoclimatology, much of which takes biological material as its basis (trees, corals, pollen, foraminiferans, diatoms etc). I used to think that claimed differences in assumptions and approaches between biologists and physical scientists were probably mostly stereotypes, but now I’m less sure about that. We biologists have, IMO, as part of our training and gestalt, a near universal appreciation for the serious, potential inferential problems caused by complexity, and I’m not convinced that the same can be said for some physical scientists. When you look at the history of tree ring analysis, which goes back about 100-120 years, you find that most of the developers of the inferential methods still in common use were in fact physical scientists interested in physical phenomena (climate and its drivers mostly), not biologists interested in physical phenomena. This matters, because their assumptions about what limits tree growth in various environments are overly simplistic, affecting both how they do their field sampling and their analysis of the data collected therefrom. There have been some exceptions though in the last 25 years, and these people have made big contributions to the field

    In your other post you asked what had caused a given change of mind. For me, that day that the clouds parted and a booming voice called out “please start getting a clue about some important concepts here Bouldin; I’m really tired of this crap and not messing around on this any more” —that had a lot to do with it, even if the jury’s still out on final effects.

Leave a Comment

Fill in your details below or click an icon to log in: Logo

You are commenting using your 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