I published my very first paper in Oikos (Fox and Smith 1997). It was part of my undergraduate honors thesis. I spent the summer before my senior year working with David Smith, Josh Van Buskirk, and another undergraduate at David’s field site off of Isle Royale in Lake Superior. It was a very rustic set-up; we stayed in a cabin on a tiny islet without electricity or running water, and we kayaked to the field site on another islet. Once a week we went to the nearest ranger station to shower and pick up groceries we’d ordered from the mainland (except one week when our order didn’t arrive, forcing us to live on canned goods for a week). David and Josh were manipulating top-down and bottom-up forces in rockpool communities. I took on the task of sampling and identifying the identifiable heterotrophic microplankton (ciliate protists and rotifers) in those pools, as well as in a number of unmanipulated pools. I identified something like 55 species in all, including a rotifer called Paracolurella which is described as ‘rare’ in Pennak’s Freshwater Invertebrates of the United States. I recall being very proud of that find. Halfway through the summer there was a massive storm which generated huge waves that scoured out all the pools involved in the experiment. This was already David and Josh’s second year of trying the experiment, but they took it in good spirits, joking about how they should try it a third time, have it come out differently again, and then write a paper about ‘heterogeneity’.
I relate this background for those readers who think I need to ask directions to find out where ‘the field’ is. 😉
Although I’d had a great time and learned a lot over the summer, I knew that not much could be gleaned from the data I had collected. So I was keen to conduct a lab experiment once I returned to school. Actually, I was keen to conduct a massive series of experiments, but David talked me down to one. I ran a straightforward competition experiment between two bacterivorous microplankton, the cilitate protist Paramecium mulimicronucleatum and the rotifer Philodina sp. Sharon Lawler, who later became a good friend, was kind enough to give me technical advice. I grew each of the two species alone, and both together in competition, for about a month, then sampled their final densities and the final densities of bacteria. I was familiar with Dave Tilman’s R* theory and thought that the species with the lowest R* for bacteria (i.e. the species able to graze the bacteria down to the lowest equilibrium density) would exclude the other.
That’s not what happened. I got exclusion, or close to it—but the winner varied. Paramecium won in two replicates, Philodina in three. The outcomes were far too different to be attributable to sampling error. Now, any community ecologist worth his salt who sees this kind of result thinks ‘Aha—priority effect!’ Except that every replicate started with the same initial conditions. It’s not as if each species won in the replicates in which it initially had an advantage. The outcome was stochastic, or maybe effectively stochastic because highly sensitive to tiny differences in initial conditions.
I had what I thought was a clever explanation for this, to do with the fact that the two species turned out to have similar R* values. But the explanation relied on the variance around the estimated mean R* values being real rather than sampling error, which I now highly doubt was the case. I’m still not sure what happened, although I have my suspicions (demographic stochasticity amplified by the system’s density-dependent feedbacks?) Other people have gotten similar results, both in protist microcosms and in other systems (e.g., Park 1948, Mertz et al. 1976, Warren and Weatherby 2006 Oikos, Kreyling et al. in press). And as far as I know, nobody’s ever really reviewed these results to see if there are any common threads. Under what conditions do small, random variations get amplified over time, and under what conditions does such variation get damped out? Seems to me that that’s a very interesting and important question for understanding ecological dynamics in a noisy, nonlinear world. There are of course various theoretical reasons why ecological systems might be noise-amplifying rather than noise-dampening (e.g., Ives et al. 2004, Vasseur and Fox 2007, Grover et al. in press), but I don’t think those ideas have received the systematic empirical attention they deserve.
Microcosms are a good system in which to study this kind of thing, because you can control and manipulate environmental noise, initial conditions, and various deterministic features of the system (e.g., equally-matched or unequal competitors). And in microcosms you can figure out (by taking replicate samples) exactly how much sampling error you have. That control, manipulability, and knowledge of sampling error allows you to really pick apart the reasons why different sources of noise do or do not get amplified, without having to rely solely on trying to statistically estimate different sources of noise. Powerful as modern statistical estimation techniques are, they have their limits (e.g., Pascual and Kareiva 1996, Knape et al. in press).
Looking back, I think the fact that my first experiment surprised me set me down the road to becoming a better scientist. If it had come out the way I’d expected, it would’ve merely postponed the day when I was forced to think about my results rather than just fitting them into the boxes defined by my a priori expectations. And the fact that the surprise occurred in protist microcosms—a system which is sometimes mistakenly thought to be too simple to surprise—was useful as well. In the field, if an experiment doesn’t come out the way you expect, it’s often for an obvious, and therefore uninteresting, reason (e.g., ‘a huge storm destroyed our study area’), or else for such a nonobvious reason that following it up becomes a daunting prospect. But in microcosms, when you get surprised, you have some hope of figuring out why. Indeed, I’m currently analyzing some experiments designed to chase down the effects of demographic stochasticity on microcosm community dynamics.
Interesting how random events set me down the path of studying…random events.
p.s. Fox and Smith 1997 has never been cited. I actually tried to cite it twice, mainly so that I wouldn’t have any uncited papers. And both times I mis-cited it (got the year wrong)! So the citations aren’t listed in Web of Science (though they do show up in Google Scholar). Apparently the Citation Gods don’t approve of cynical self-citations. 😉