In a 2001 paper in Nature, Michel Loreau and Andy Hector described a new method, based on the Price equation, to partition the “selection effect” and the “complementarity effect” of biodiversity on ecosystem function, and demonstrated its use on data from the BIODEPTH experiments (BIODiversity and Ecological Processes in Terrestrial Herbaceous ecosystems: experimental manipulations of plant communities). Fifteen years after the paper was published, I spoke to Michel Loreau about the development of this method, his collaboration with Andy Hector, and what we have learnt since about the roles of selection and complementarity in the biodiversity – ecosystem function relationship.
Citation: Loreau, M., & Hector, A. (2001). Partitioning selection and complementarity in biodiversity experiments. Nature, 412(6842), 72.
Date of interview: 5th August 2016 (on Skype)
Hari Sridhar: What was your motivation to do this specific piece of work? It comes a couple of years after the Hector et al. paper which analysed the experiments from the BIODEPTH study. Did that paper serve as a motivation to do this analysis?
Michel Loreau: No, my motivation was much broader than that. At that time, there was a big controversy internationally over the interpretation of biodiversity experiments. One of the first such experiments was the experiment in the Ecotron that you mentioned in your email, by Shahid [Naeem] and others. Then there were experiments by Dave Tilman at Cedar Creek and the BIODEPTH project with our Science paper in 1999. All these papers elicited a lot of debate, in particular about the kind of interpretation we can gain from these results. Some people became very critical about these experiments, saying that the results only showed that, on average, plant productivity increases with the number of species. For them, that was no proof that there was a true diversity effect because it could simply be a statistical effect of having sampled more productive species. So looking at the average productivity at each diversity level was not informative. The debate became so strong that it went to the public newspapers and became aggressive among a few people. Science is not always purely objective; there are also personal components. Anyway, at that time I was working within the BIODEPTH project and I was quite open minded when I began. I was not sure that we would actually find diversity effects systematically. The first model I published on this issue was in 1998, showing that it’s not at all inevitable to get that result and you could have the opposite result as well. Then I organised a big conference in 2000 in Paris to bring everyone together and try to resolve the controversy. We published another famous paper in Science in 2001, where we present a consensus view about what can happen in biodiversity experiments. Our conclusion was that two types of effects can occur at the same time: complementarity effects and selection effects (other people call them sampling effects but in my opinion this terminology is a bit confusing). At the same time, I developed a new approach that tried to separate these two effects. First I made it clear why I think the term “selection effect” is more appropriate than “sampling effect”. This came with the realisation that I could actually use some equations coming from evolutionary theory – the Price equation – which describes changes in gene frequencies in populations. In this equation, there is a selection term based on a covariance. I had the idea to apply the same approach to ecological data in a different setting to see if there was a covariance between the intrinsic productivity of species in monoculture and their dominance in mixture, because that, as I showed, was the real basis of the “sampling effect” argument. The argument was not about sampling per se, but about the fact that productive species would dominate mixtures. That’s where this came from, so it’s not just from the BIODEPTH experiment. It was a huge debate and there was a big need at that time to find new ways to tease apart the various hypotheses that were proposed to explain the results of biodiversity experiments. At the same time, I tried to build a consensus on these issues in the scientific community. That’s why there were two papers at roughly the same time – the review paper in Science in 2001 and this new approach to biodiversity effects in Nature.
HS: Stepping back a bit, could you tell us how you got interested in this topic? Looking at your publication profile, I see that a lot of your early work was on carabid beetles, and it’s around 1996 – correct me if I’m wrong – when you started working on biodiversity and ecosystem functioning. How did this interest come about at this time?
ML: That’s an interesting story. I have not had a straight career in science. I first started as a young research scientist working on carabid beetles, indeed for quite a long time. At that time, my professional situation was rather difficult. I didn’t have a permanent position until I was 39, which is very late. At some point, I decided to leave the university and do something else. I was hired as a programme manager at the Science Policy Office in Belgium from 1990 to 1992. I was in charge of the global change research programmes in Belgium. This was an interesting experience that allowed me to step back from my own research and to get interested in much broader issues. I decided to change my research topic at that time. I anticipated that the next frontier in ecological science would be the relationship between biodiversity and ecosystem functioning, and decided to work on this topic. At that time, biodiversity and ecosystem functioning was not yet recognised as a big issue, but some ideas were starting to pop up. I decided that this issue had to be studied properly from a theoretical perspective. When I went back to the university in 1992, I started to build ecosystem models as a first step. Then I focused on biodiversity and ecosystem functioning, and published the first ever theory paper on this topic in 1996. That was a big change in my career path, which resulted from a conscious choice that I made at that time. I then met some of the big names working on this topic from an experimental angle, in particular John Lawton. John is the person who imagined the Ecotron and brought Shahid [Naeem] and others over to perform these experiments. I started to collaborate with them and applied my theory to experimental data. But I had made my decision to work on this topic before any experimental study was published.
HS: Do you remember when you first encountered the Price equation, and when you realised its utility in the context of biodiversity and ecosystem function?
ML: I don’t have a precise date or year. If you look at my list of publications you will see that I have also worked extensively on spatial ecology and evolutionary ecology. In particular, I developed some new models of evolution in ecosystems. The idea was to look at evolution, not only as a property of a species, but also as an emerging property in whole ecosystems. I published a number of papers on that topic around 1998-2000 with my colleague Claire de Mazancourt, who at that time was my PhD student, and that’s when I discovered the Price equation. I was interested in evolutionary issues and the Price equation was one way to formalise them. That gave me the idea to use it in another context.
HS: How did your collaboration with Andy Hector come about? I notice that, before this paper, both of you had independently worked on theoretical aspects related to separating out the sampling and complementarity effects. How did you decide to work together?
ML: Andy and I were both involved in the BIODEPTH project, which led to that Science paper in 1999. We had a lot of discussions within the project. In fact, there was a lot of debate about the interpretation of experimental results in that project. It wasn’t like a monolithic group of people. It was quite intense, you know, but friendly. It was very stimulating. Andy and I were both interested from the beginning in trying to use relative yield approaches that come from agricultural sciences to quantify biodiversity effects. We actually published a debate between the two of us in Oikos in 1998, in which Andy basically argued that relative yield approaches should be used in biodiversity experiments. My response or complement to his paper was: that’s a good idea but there are many ways to do that. I showed the strengths and limitations of different types of approaches. When I developed the new method based on the Price equation in 1999, I immediately contacted Andy because he was interested in the topic and he had access to the data from BIODEPTH. I felt that just developing a new approach was not enough; we needed to apply it to real data. Since we both shared this interest, it was quite easy to collaborate on this. He analysed all the data and I basically provided the theory behind it. It was a very fruitful collaboration. Since then we have collaborated on many occasions, and we continue to do so today.
HS: It looks like you were at a different university at that time. Were you a professor at that university?
ML: Yes, I was a professor at Pierre and Marie Curie University in Paris. Then I moved to McGill University in Montreal before coming back to France on my current position.
HS: How did this collaboration with Andy work? Did you meet often to discuss work or was it mostly done over email and phone?
ML: It was mostly done through email. As I said, we had already discussed a lot in the BIODEPTH and other meetings. We had at least one yearly meeting with everyone present in BIODEPTH. I also organised or participated in a few workshops at Silwood Park where Andy was working, and that conference in Paris. So we saw each other on a regular basis, even if not specifically about this paper. With regard to this paper, once I got the idea I shared it with him and he was enthusiastic. So it was not difficult to organise this work from a distance. In a way it was the crystallisation of something we had both been interested in, and it was easy to collaborate on that paper.
HS: Do you remember how long it took from idea to publication?
ML: I had the idea in September of 1999. I remember exactly when and where. It took some time until I discovered that it was really working. It took me a few months to try with examples, and then I contacted Andy, and it took us about a year to do the analysis, and half a year to write a manuscript. So two years in total, from idea to publication. It’s not that long for a paper like this.
HS: You have made me curious! I want to know more about this Eureka moment. When and where did it happen?
ML: It was on a boat in Greece. I was invited to give, as far as I remember, my first keynote lecture in an international conference. That was 8th European Ecological Congress (EURECO ’99), which took place in Porto Carras. I was invited at that conference to talk about biodiversity and ecosystem functioning, and while we were on a boat trip sightseeing, I had that sudden Eureka moment. I knew the Price question and I had been thinking about using it but I hadn’t found the right way to do it so far. Then there was the relative yield approach that Andy and I had discussed about previously. The two came together suddenly and I just found the equation, not on a piece of paper, but in my head.
HS: Was this before or after you gave the keynote talk?
ML: Ah, that’s a good question. I am afraid I don’t remember.
HS: I was wondering whether the idea came before, and you included it in your talk.
ML: No, I didn’t.
HS: Did you do most of the writing for the paper?
ML: I don’t remember exactly how it worked, but since I’m the first author and I developed the theory, I wrote the first draft, and Andy wrote the parts on the data analysis. It was a true collaborative effort. It’s always been like that with Andy; we collaborate so well that we don’t even have to discuss much. It’s almost spontaneous.
HS: Did you have a writing routine at that time – a time and place when you wrote?
ML: There was no particular time or place. When I was in Paris for 11 years, I was commuting between Brussels and Paris often, because my family was still living in Brussels. I was in Brussels for half of the week and the other half in Paris. So I was used to working at home in Paris, at home in Brussels, in the train etc. I was really not attached to any particular place and time.
HS: Did this paper have a smooth ride through peer-review? Was Nature the first place you submitted it to?
ML: Yes, that one was quite easy. We did receive some comments, but they were mostly enthusiastic, so the ride was rather smooth.
HS: How was the paper received when it was published, in the popular press and in academia?
ML: It mostly received a lot of attention from people working in the field. Scientists, basically. I don’t remember getting a lot of attention in the popular press. My colleagues who were working in the area discovered, quickly, that this new approach was quite useful, and so many people started to use it in subsequent experimental work.
HS: This paper has been cited over 1000 times. Would you know if most of the citations are by studies that use your method?
ML: Yes, at least in the beginning. Nowadays it’s used more broadly by people interested in the field, but I believe it is indeed mostly cited by people using the method, or discussing it at least, even if they don’t use it directly. For example, discussing that this could be done if we had the right data. It’s also cited for its conceptual basis, the concepts of complementarity and selection.
HS: I would like spend a few minutes on the Acknowledgements, to find out more about how you knew these people and how they helped.
ML: Sure. Shigeo Yachi was my post-doc in Paris from 1997 to 1999. We discussed a lot about the Price equation together. We both had the idea that it would be good to use it in the context of biodiversity and ecosystem functioning experiments. I found out the right way to do so but he helped me build the idea and make the right connections, so I felt that he deserved to be a co-author on the paper. I invited him to be a co-author but he declined because he felt that he didn’t contribute enough. You can’t force anyone to be an author but I feel he played an important role in the concept.
Bernhard Schmid and Doug Deutschman helped us dealing with data analysis. Bernhard was also part of the BIODEPTH project. He was in a way the statistical “head” of the BIODEPTH project, and so we discussed with him about how to best analyse the data. I believe the same holds for Doug Deutschman. I don’t remember exactly because I’ve never met him personally, but I think he was contacted by Andy to provide feedback on data analysis.
HS: You also thank other members of the BIODEPTH project. Did they provide inputs specifically on this paper or is the acknowledgement for participation in the larger project?
ML: Mainly because they collected the data that we used. I think it’s quite normal to thank them for that because they are not co-authors on the paper.
HS: Did they also provide inputs specifically on this analysis and drafts of this paper, or was that just you and Andy?
ML: No, that was just Andy and me. Otherwise they would have been named in the Acknowledgements.
HS: The grants that you acknowledge – were they obtained specifically for this paper or for the larger project?
ML: Both those grants were for broader projects, but they helped me to get a bit of funding to travel. The PICS was a collaborative project with the UK that helped me travel to the UK and had people travelling from the UK to France. It helped to organise some meetings but it was not at all specific to this project.
HS: Did this paper have a large impact on your career and your future research?
ML: Yes and no. It was part of a large series of papers on this topic and I don’t think that any specific paper stands out. Perhaps the Science review paper was more influential in a way because it’s less methodological, it can be read by anyone and so it has been more cited because of that. It can be used by policy makers, conservationists, anyone who is interested in the topic. The Nature paper is a research paper, so you have to be interested in biodiversity experiments to read it. But it is definitely one of my favourite and most influential research papers, and it did help my career.
HS: It is now 15 years since the paper was published. Would you say that the main findings, with regard to the relative importance complementarity and selection for this set of sites, still hold true more-or-less? Or has additional data changed the result in anyway?
ML: Yes, its main findings have been remarkably robust. For instance, we had another debate after the first one, with Brad Cardinale and others who published a review paper in Nature in 2006. Their paper suggested again that biodiversity effects in experiments were largely driven by the sampling effect. Andy and I didn’t agree on their conclusions, but instead of fighting against each other, we decided to work together to perform a meta-analysis of all the grassland experiments that had the appropriate data. That joint work was published in PNAS in 2007. Andy and I were convinced that complementarity was important even in their data based on our previous experience. Brad and others were convinced that this was not the case. So we decided to analyse the data together using our method. It took one year for Brad to accept that we were right but in the end he did, which I view as a valuable sign of scientific honesty. Now we have more and more examples showing that complementarity is the leading factor in biodiversity experiments, so the conclusions of our paper hold more than ever. Our method does not solve everything, it has some limitations, some of which are even mentioned in the paper. But overall I would say that both the method and the results of our paper have stood the test of time particularly well.
HS: You said that even in this paper you identify some of the limitations of this method. Did you work on these limitations after this paper to try and refine the method?
ML: Other people have tried to do so. We provided an additive partitioning with two components, because we wanted it to be as simple as possible and as transparent as possible. Jeremy Fox and others have developed our approach to build equations with three or more components. Although mathematically correct, I don’t feel these more complex equations have led to major new discoveries so far, because the new components don’t always have a clear interpretation.
To be honest, I was not especially interested in developing the method. It’s not a kind of magical recipe that will solve all problems. I have never believed in that kind of thing. I prefer to concentrate on new questions and not just to refine one particular approach. But that’s my bias. I know some people like to continue building what they have done. I’m different. Once I solve a problem I tend to move to another problem. That’s my way of doing science.
HS: Is the method still standard practice in studies trying to separate out the effects of complementarity and selection?
HS: You say “the experiment was not designed to fully separate the effects of legumes or functional-group diversity from that of species diversity”. Subsequent to this paper, were there studies that tried to do this?
ML: Yes indeed. In particular, Jasper van Ruijven and Frank Berendse published the results of a very nice experiment testing for the effects of legumes in 2003. What is interesting is that Frank was critical of previous papers showing effects of biodiversity on ecosystem functioning. He didn’t believe that biodiversity had such a big role, and he thought that the effects showed so far were due to the presence of legumes. So Jasper and he designed a clever experiment excluding all legumes to test their hypothesis that biodiversity should have no effect in the absence of legumes. To their big surprise, they found exactly the same positive effects of biodiversity without any legume. It looks almost like a fractal process. You first have the legume-grass complementarity but then within each group, grasses or legumes, you still find the same kind of complementarity. A kind of Russian dolls scenario.
HS: In Box 1 of your paper you say “Our approach, however, could be applied to other organisms and other experimental designs in which density is allowed to reach natural levels. As it is based on variation in relative yield (DRY), it could also be used to study how biodiversity effects change through time”. Has there been research along these lines?
ML: I’ve done that with the Cedar Creek data. We published another paper in 2007 with the Cedar Creek data looking at how selection and complementarity change through time.
HS: In the last sentence of your paper you say “It [your method] cannot replace direct experimental investigations into the mechanisms at work in responses to biodiversity changes at the ecosystem level, which are now critical to further progress in this area.” Have there been experiments that directly test these mechanisms?
ML: Yes indeed, but not that many. Surprisingly, the number of such experiments is relatively limited compared with the field as a whole. There have been maybe a few dozens of experiments testing specific mechanisms. And what people realised based on these experiments is that complementarity is really a generic term. Complementarity can occur through very different biological mechanisms. It can occur through different rooting systems, i.e. plant roots access different nutrients in the soil at different levels. It can also happen through different foliage architectures, which increase light acquisition, and even through different soil pathogens. It seems that the general concept is particularly robust but that the details of the biological mechanisms involved can differ quite a lot. The same is true for the selection effect. Selection can be driven by many traits and can also vary quite a lot. So I see our method more as a framework to test hypotheses rather than to identify biological mechanisms. It can give insights into what is important but if you want to understand better then you have to go into the details of the mechanism that produce these effects.
HS: Have you ever read the paper after it was published?
ML: Sometimes I do go back to it, in particular to the Box. Someone asks me a question: I don’t find the same results, why is that? Then I go back and look – ok, the method was correct, then what is wrong with the new analyses? Sometimes I go back to specific sentences where we define complementarity and selection because there has been so much debate on these concepts. Not deep debate, but people asking what you really mean by complementarity and selection. It’s important to go back to the sentences and look at exactly what we wrote. Andy and I and other authors actually wrote other papers to clarify what we mean by these effects and, in particular, how they differ from biological mechanisms.
HS: Do you think your writing style has changed today?
ML: Not really, I don’t think that it has changed in any major way. To be honest, I have fewer opportunities to write pure research papers on my own nowadays. Most of the time, I develop new ideas and help my students and post-docs to formalise and publish them. I write more synthesis papers and things like that. But I still write in a similar style as in this paper.
HS: Would you count this as one of your favourite papers, among all the papers you have written?
ML: Yes and no. I don’t have a single favourite paper but this one is clearly one of them. I like the fact that it was a nice idea that came spontaneously. And I believe it did help to unlock the field. I think this paper had a bigger impact than just its number of citations because the whole field of biodiversity and ecosystem functioning might have collapsed if it hadn’t been published. Not because I’m particularly influential but simply because there was so much debate at that time, and this paper provided a way to move beyond the controversy. I believe it helped a number of people to remain in the field because they saw some light at the end of the tunnel of that debate. I also like it because it’s clear, it combines theory and data, and that’s the kind of theory I like. But I also have other favourite papers for different reasons.
HS: What would you say to a student who is about to read this paper today? What should he or she take away from it?
ML: To be honest, I am not sure. Perhaps I would emphasise the context in which it was written. People tend to forget history very easily. It’s important to understand history, to see the context in which science develops. In ecology there is this tendency: people start working in an area, there is a debate without resolution, and the area collapses. Twenty years later someone starts exactly the same thing and the cycle repeats. There’s little memory. For me, memory is important.
I view this paper as one among others that are still useful today. So I would submit it among other papers to read in that field, but it doesn’t have a special status. For me, personally, it was an important paper, one of my favourites, but to my students it’s just one among other papers to read if they work in that area.