I was invited to give an Ignite talk at the ESA meeting this year. I figured it would be fun to try, so I said yes. But now I need to actually start writing it, and that’s where you come in.
The session was proposed as a debate on theory vs. empiricism in ecology. The speakers are even supposed to be divided into “Team Theory” and “Team Data”. I’m not sure if there will actually be any fireworks, though, since everybody involved believes in the value of both theory and data. Indeed, several of the people involved, including me, do both modeling and data collection.* Heck, people on both “sides” have even co-authored papers with one another! So don’t come expecting a fight (though you never know…), but do come expecting thought-provoking comments on how ecologists use theory and data, and how we can use them better.
That’s my goal, anyway: say something sufficiently provocative that people will want to discuss it and quite possibly disagree with it, but not so outlandish that people will just roll their eyes and dismiss it. Ignite sessions are pretty informal, they are to regular talks as blog posts are to peer-reviewed papers. Like blog posts, they’re a space in which it’s ok to push boundaries, try ideas on for size, be a little playful, and say things that are worth saying but that you maybe couldn’t say in a more formal setting.
Which is where you come in. I’d love to have suggestions on what to talk about.** I have some ideas of my own, of course, but they’re tentative:
- I’m on Team Theory, so I have to say something extolling the value of theory.
- My tentative plan is to address the question “When does data settle arguments, and when does it not?” And my tentative answer is “Only in limited circumstances, where the question of interest is precisely-specified and can be addressed with rather simple empirical studies.” The inspiration comes from some old posts. For instance, I think the questions “Is interspecific competition common”, “Are trophic cascades common?”, “Is population growth typically density-dependent”, and “What is the relationship between plant species richness and total plant biomass, all else being equal?” were all settled by empirical data. People went out and did lots of competitor removal experiments, lots of predator removal experiments, lots of appropriate statistical analyses of long-term time series data, and lots of “random draws” BDEF experiments, and the accumulated data answered those questions beyond any reasonable doubt. (For anyone who thinks I’m rewriting history here: now would be a good time to speak up! 🙂 )
- But while it’s great when data can settle arguments in this way, I think such happy situations are rare. Think of various famous historical cases in ecology where big empirical research programs failed or ran into problems due to lack of theoretical grounding, grounding in mistaken theory, or grounding in misunderstandings of correct theory. So even though lots of data were collected, questions weren’t answered, and if anything arguments were started rather than settled. I’m thinking here of the IBP, early work on trying to infer competition from observational niche overlap data or species x site matrices, the idea that species interactions are more specialized and intense in the tropics, and (of course) the intermediate disturbance hypothesis. So this is where the importance of theory comes in.
- I don’t want to to just list some cases where weak links to theory led to problems. That would be boring, I think. I want to diagnose the problems, say something about how they arose and how to avoid them. As I’m sure is obvious from the rather heterogeneous lists of historical case studies I just gave, I haven’t gotten too far on the “diagnosis” part. In particular, I’d like to be able to say something about when going out and collecting data helps us recognize serious problems with the questions we’re asking. You’d think that’s something that would happen often, but I’m not sure that it does. Thoughts?
- On the presentation side, I’m thinking of not having any slides at all. Or just having every slide be the same pretty (or silly) picture. Given what I’m talking about (a subject that doesn’t have many obvious visuals, though I’m sure I could come up with something), and the fact that I’m only talking for 5 minutes, I’m not sure that slides would add much. They might even be a distraction. Maybe better to just force people to listen to me for 5 minutes, rather than having them try to listen to me and read slides that are only up for 20 seconds at a time? Indeed, I’m thinking that the novelty of a slide-free talk might itself actually be a more effective way of engaging the audience than whatever I might put on my slides. Or maybe I’m trying to be too clever here (or too lazy, since it’ll be easier to prep my talk if I don’t bother with slides)?
- An alternative plan: do something much narrower and less ambitious, and just focus on the zombie IDH as a case study of the importance of having good theory and using it to ground your empirical studies. Pros: easy to prepare, involves zombie jokes. Cons: I’d just be saying what I’ve already said in numerous blog posts and a peer reviewed paper, which seems likely to bore or disappoint the audience, and would suggest a lack of ambition on my part.
So that’s what I’m thinking. What do you think? If you were in my shoes, what would you say and how would you say it? What examples would you use? What visuals, if any? Fire away–I’m totally open to suggestions!
*I kidded the organizers about this when they first told me the tentative invitee list, joking that they apparently didn’t know any people who disagree with one another when it comes to theory vs. empiricism. But in seriousness, I’m sure it’s better this way. As Susan Harrison noted in an old post on debates at scientific meetings, participating in a debate is no fun if you can’t respect your opponents and enjoy hearing their arguments. It’s difficult to have that respect and enjoyment if people’s world views are too different.
**For those of you who are thinking, “Wait, you had to submit an abstract months ago, so don’t you already know what you’re going to say?”: umm, no. My ESA abstracts are like religious texts. They shouldn’t be interpreted literally. 🙂 When I write my abstract, I know what topic I’ll be talking about, but I have to guess what I’ll say about it. “Jeremy Fox ESA Abstract Fundamentalism” is not a religion you should adhere to. 🙂
Dynamic Ecology 
Maybe you can take the claim that data are often “theory-laden” as your point of departure. Both whether a certain type of data is being collected in the first place and how existing data are being interpreted often strongly depends on theory. For example, what sex allocation research was being undertaken and what were researchers making of biased sex ratios before Hamilton’s extraordinary sex ratios (1967) and after?
Theories, on the other hand, are often quite inert to data. That’s why a certain kind of interpreting Popper is called naive falsificationism. One nasty fact may kill a hypothesis (prediction), but not a whole theory because ad hoc assumptions can be made to fit the data. The ad hoc assumptions are not always bad. Often they even turn out to be correct, especially, when one nasty fact tries to fell a honed theory (e.g., faster than light particles vs. broken instruments).
If you complement the above with fitting examples from ecology and defend the thesis that theory is in the driver seat, I guess, you’ll get a lot of opposition from field ecologists.
Thanks Joachim, this is helpful.
Hi Jeremy, as you might have guessed I disagree pretty strongly with your statement,
“Only in limited circumstances, where the question of interest is precisely-specified and can be addressed with rather simple empirical studies.”
And I would guess that this is where the theory versus data argument gets polarized – not on whether to use data or theory but on “When does data settle the argument?” I would say that ONLY data can settle the argument and until theory is tested with data it is simply a hypothesis. But, I would probably take it a step further and say that the argument doesn’t get settled until you test the theory on data that weren’t used to construct the theory. Now, I get that theories that have been parameterized using data may be a step beyond ‘just a hypothesis’ but the degree which such a theory is better than ‘just a hypothesis’ is to the extent that data have allowed it to be realistically paramaterized.
Wish I could be at this discussion.
And as an aside – what is the answer to “Is population growth typically density-dependent?”?
Best, Jeff H
@Jeff:
Thanks for your comments Jeff. Although you won’t be there, hopefully some folks who think as you do will be. The whole point is to spark a vigorous discussion.
“Is population growth typically density-dependent?”
Yes (though given the short length of most ecological time series, our estimates of density dependence often have wide confidence intervals and aren’t significantly different from zero)
I don’t hink that theories are just graduated hypothesis. A hypothesis is a prediction that can be tested (falsified), whereas a theory is a(n established) set of ideas that can be used to make predictions.
The best way to illustrate the difference is probably by analogy. Let a theory be analogous to a model and a hypothesis be analogous to a prediction drawn from that model. The prediction can be tested and falsified, but the model will not therefore be automatically discarded. One can fiddle with parameters etc. and keep the model (theory) in a slightly modified (improved) form.
Here, data come second in correcting or improving a model (theory).
For me, the real value of theory is identifying what processes are really important for answering your questions. A question I am often interested in is “How do variable environments affect interspecific competition?” An empiricist might think of ways to manipulate environmental variation and measure some aspect of competition between species. But I would immediately wonder, what should I measure? I might measure three aspects of competition between two species and see that environmental variation does different things to those three aspects of competition (say growth effects between adults, effects on reproduction, and effects on juvenile survival). What am I to say? Without any guidance, I might be left to say that the effect varies based on system – which (at least personally) isn’t very satisfying.
Luckily for me, I think we have some good theory to help one work through such outcomes. I would argue that the guidance provided by such a theory paves a way for designing experiments and answering questions in a much more efficient way than without.
It may be something of a tangent, but I wonder if that Prime Directive of Data Analysis rules (“don’t extrapolate beyond the bounds of your data”) could be of interest. The Directive may be fine when talking about fitting a mathematical model (e.g., regression) to data that have been collected, but one could argue that the real purpose to having a model is to create a framework where it is acceptable to extrapolate. If we were going to restrict ourselves to the bounds of our data in the strictest sense, then we can really only talk about the spatial, temporal, and other dimensions (e.g., analyte concentration) from that data set. Without a model (either mathematical or conceptual), we can’t extrapolate to other times, to other places, to non-measured inter-relationships… Experimental data are how we validate a model, and maybe even how we formulated the model to begin with, but there really isn’t anywhere to go until you start modeling.
Pingback: Friday links: zombie ideas in evolution and psychology, a world without statistics, and more | Dynamic Ecology