When faced with a set of items that can vary along multiple axes or dimensions, it’s often interesting to locate the items in a multidimensional space. Not just to see where they fall, but to see where they don’t–where are the “holes” in multidimensional space? Ecologists do this with species composition data, looking for “forbidden combinations” of species that perhaps can’t coexist because of competition. They do it with morphological data, looking for unoccupied areas of “morphospace” that evolution has failed to fill, possibly because of character displacement, or because of constraints that make it hard to evolve certain combinations of morphological features. And we can do the same thing with our own research!
Steve Walker gave me this idea, in his fun old post classifying different philosophies of statistics in a four-dimensional space. I’ve been wanting to apply the same idea to ecological research projects for a while now, just for fun but also to provide a basis for thinking about serious issues like how the field of ecology has moved through “research space” over time. I’m going to stick to four dimensions, because that seems like enough to be interesting, but not so many as to be overwhelming. So here are four dimensions along which empirical ecological research projects can vary:
- Is the research conducted by an individual or small group, or by a large, coordinated team?
- Is this research question- or hypothesis-driven, or not?
- Is the research based on observational data, or data from manipulative experiments?
- Does the research involve collecting new data, or is it based on existing data?
I could’ve added a fifth dimension for whether the research is empirical, theoretical, or a mix. And perhaps a sixth dimension for whether the research involves expensive equipment or not (think, e.g., whole-genome sequencing or high-tech remote sensing, vs. the low-tech gear used in much standard field ecology). And I could’ve added dimensions for spatial and temporal grain and extent. But like I said, four axes seemed like enough as a starting point. Feel free to suggest other dimensions in the comments.
To keep things simple, I’m going to pretend that there are only two possible locations on each axis. I don’t think much would be gained (in terms of either fun, or actual insight) by trying to be more quantitative than that right now. So I’m proposing a four-dimensional space with a total of 2^4=16 possible locations.
Here’s where I think various sorts of ecological research fall in that space:
- A typical single-investigator grant from an agency like NSF, NSERC, or NERC involves an individual investigator, doing hypothesis-driven research (hard to get funded without hypotheses!), and collecting new data, either experimental, observational, or both.
- NutNet is a coordinated team, doing hypothesis-driven research and collecting both new experimental and observational data.
- An NCEAS working group is a coordinated team, doing hypothesis-driven research (at least usually), based on existing observational or experimental data
- NEON comprises coordinated teams, doing hypothesis-free research based on collecting new observational data. Other large monitoring efforts like the Breeding Bird Survey, the Christmas Bird Counts, and coordinated citizen science efforts to monitor things like flowering phenology also fall into this category. Further back, the IBP basically fell into this category.
- LTER sites are a bit difficult to classify, because in some ways they can be thought of as single (complex) research projects, and in other ways they can be thought of as a bunch of separate research projects that all just happen to occur at the same location. I guess I’d classify them as coordinated teams, conducting both hypothesis-driven and hypothesis-free research (LTER sites host experiments, but also engage in monitoring), by collecting new observational and experimental data.
- An individual who writes a review paper is an individual using existing data to pursue research that might be either hypothesis-driven or not, and based on either experimental or observational data. I do think all combinations are possible here, although hypothesis-free reviews are probably fairly rare these days, and hypothesis-free reviews of experimental data are a bit difficult to imagine.
If you’re scoring at home, the above examples (which are merely the first ones that occurred to me) collectively cover 12 of the 16 locations in our four-dimensional space. The four gaps are as follows:
- Individuals doing hypothesis-free research based on collecting new observational data. This is basically the sort of work with which ecology began–think of amateur naturalists in Darwin’s day, just observing and collecting the local flora and fauna. I’m sure there probably are individuals still doing this sort of research. But because of the inherent limits to how much data a single individual or small group can collect, this sort of work doesn’t figure very prominently in modern ecology.
- Individuals doing hypothesis-free research based on collecting new experimental data. Pretty rare for people to do totally hypothesis-free experiments, just “kicking the system to see who yells”, as my undergrad advisor once put it. And quite rightly–this usually isn’t a very good way to do science, although there are limited contexts in which it can be useful.
- Coordinated teams doing hypothesis-free research based on existing observational or experimental data. Have there been working groups, at NCEAS or elsewhere, just doing purely descriptive work? Or maybe exploratory working groups could be put in this category?
Wasn’t sure what I’d find when I started to think about this. But I think it turned out kind of neat. While the 12 “occupied” points I identified aren’t all equally occupied (as I noted, hypothesis-free research by individuals, based on existing data, is probably rare), I do think that they’re mostly more occupied than the four “gaps” I identified. And it seems to me that the four gaps exist for good reasons. They’re the four research approaches of the most limited scientific value, I think. In general, it seems like the gaps, and the more lightly-occupied parts of the space, are the “hypothesis-free” parts. I also think it’s interesting that my four dimensions appear to be largely orthogonal. It’s not that, say, large coordinated teams only fall in one or two locations on the other axes or anything like that. And while I’m sure the relative occupancy of different points in this space has varied over time, I don’t think there are any occupied points that until recently were complete gaps.
What do you make of this?