In a paper published in Ecology in 1994, Timothy Wootton demonstrated the use of a new approach that combines experimental perturbations with path analysis to determine interaction strengths in communities. Wootton showed, using interactions between birds and other members of an intertidal community, that this approach could be an efficient way to predict direct and indirect interactions among unmanipulated species withing communities. Twenty-three years after the paper was published, I asked Tim Wootton about his motivation to do this study, his memories of field work and what we have learnt since about working out interaction networks in communities.
Citation: Wootton, J. T. (1994). Predicting direct and indirect effects: an integrated approach using experiments and path analysis. Ecology, 75(1), 151-165.
Date of interview: Questions sent by email on 29 December 2017; responses received by email on 21 June 2018
Hari Sridhar: I would like to start by asking you about the origins of the study presented in this paper. This study seems to have formed a part of your PhD and you even cite a couple of earlier papers of yours on indirect effects. How did you come up with the idea of linking field experiments, natural history observations and path analysis in this paper?
Tim Wootton: This was part of my PhD. It developed out of a question a colleague (A. Richard Palmer) asked about whether birds had an impact on predatory dogwhelk snails. This question was based on an observation that there are bright orange morphs of these species which are very rare, yet the gene controlling this morph is dominant, suggesting strong natural selection against the morph. Yet when I tested for effects of birds in my experiments, I found an unexpectedly strong positive effect on the snails. In thinking about the natural history and web of species interactions in the system, I was able to imagine several scenarios by which this might have arisen through indirect effects. Construction of such post-hoc stories was the standard way of inferring indirect effects. I was interested in whether there might be a more objective way to evaluate hypotheses of indirect effects. Doug Schemske’s plant evolutionary genetics lab had been playing around with path analysis (first invented by Sewall Wright for population genetics problems) and in talking with his colleagues, it struck me that this method was worth applying to webs of species interactions.
HS: Stepping back a bit, how did you decide to do a PhD in ecology? I notice that some of your early papers are on birds. Did your interest in ecological research start from an interest in bird watching?
TW: I grew up in rural New York state outside of Ithaca (I was a Cornell “faculty brat”), and was outdoors all the time. In this environment, I developed an early interest/obsession in birds–my parents tell a story of me identifying “wing-wing blackbirds” out of the window of a moving car at age 2. I maintained this interest throughout my childhood, and began collecting bird nesting data for the Cornell Laboratory of Ornithology in junior high school. Also, on a family trip to Florida, my mother directed my interest toward collecting shells to get me out of her hair, and I continued to semi-obsessively collect shells and observe marine life whenever I had a chance to visit the shore. Like many kids at the time, I was also influenced by the television show “Wild Kingdom“, and my parents subscribed to National Wildlife magazine. From these outlets I developed an interest in species extinction, an emerging concern at the time. My parents took the family on sabbatical to Stanford University during my first year of high school, which resulted in many visits to the California shore. I read as much as I could about the plants and animals I was seeing on the shore, and was particularly influenced by Rickett et al.’s Between Pacific Tides with its very modern emphasis on species interactions with each other and their environment. In particular, during my reading about shoreline ecology, I came across a description of a study describing how some guy removed starfish from the shore, which caused the collapse of associated species. This struck me as a most compelling way to understand the implications of species extinction.
HS: If you don’t mind my asking, how come your PhD supervisor was not an author on this paper?
TW: When applying to graduate school, I was fortunate enough to stumble across “that guy who did the starfish study”, Bob Paine at the University of Washington, and made it into his lab. Bob was a strong believer in students maintaining an independent intellectual identity as a way both to increase the marketability of his students upon graduation and to increase the intellectual and methodological breadth and perspective of his lab. Therefore, he actively discouraged students from working directly on any of his own projects and did not engage in slapping his name on his students’ papers, so he published very few papers with his students. I try to maintain this approach in my own lab, as Bob was remarkably successful at producing graduates who went on to faculty positions of their own.
HS: How did you choose your field site for this study? If you think back to the days when you did fieldwork for this study on Simon’s Landing site on Tatoosh Island, what are your most striking memories? What was a typical day like? Did you work alone or did you have people to help you with fieldwork?
TW: Tatoosh Island, which is owned by the Makah Tribe, was Bob Paine’s primary study site, and he encouraged his students to explore possible projects there. It took no salesmanship on his part to convince me to work on Tatoosh. Because of the cool moist environment, the timing of low tide events relative to stressful air conditions, the high wave action, and the minimal direct human impact, the species diversity and abundance on Tatoosh is spectacular compared to other shores that I have visited, and the isolation of the island ensures that equipment and experiments are not disturbed by curious visitors, so it is an ideal site to do research. My first visit to the island in November 1984 is probably still the most memorable. I got to fly out by helicopter, and once there I was awed by all the eagles, peregrine falcons and marine mammals I saw, along with all the shore life, and the trip was punctuated by the appearance of a dead humpback whale on the beach, which had decayed enough internally that its bones were disgorged when the tide washed it back out to sea the next day.
I had a cursory introduction to Simon’s Landing (our informal name of the site after a famous theoretical ecologist fell from the top of the overlooking cliff and nearly killed himself; I have not learned of any formal name of the site from the indegenous Makah Tribe who own the island) from afar on my first visits, but little work was being done there at the time, because of the difficult access down/up the cliff to get to it, and the winter tides on my first visits are at night, so it was hard to see what was there. When taking classes at Friday Harbor Laboratories the following spring, I saw the broadcast of the shoreline episode of David Attenborough’s Living Planet series. Although there was no cable TV available and viewing the segment was like looking through a violent blizzard because of poor antenna reception, when the segment on rocky shorelines appeared I could make out this fantastic uniform sloping rock bench with strong patterns of zonation that the show was using as its representative rocky shore, and I said to myself “wow, that spot would be a really great study site!” That site turned out to be Simon’s Landing, and I turned out to be right.
We typically visit Tatoosh, which is only accessable with permission of the Makah Tribe, for 4-5 days every 2 weeks between April and September. For most trips, we land through the surf in a small zodiac boat, and live out of two small buildings that lack running water and have minimal power provided by a solar panel. Summer tides are in the early morning, so we wake up around 5 am, work until around noon, then do research and camp maintenance (clearing trails, preparing equipment, transcribing data, read literature, review manuscripts, etc.) for the rest of the day, all the while keeping our eyes and ears open for interesting birds and other aspects of nature. While most research projects are done alone or with the help of an undergraduate field assistant, all members of the research group (1-2 Principal Investigators, a couple of graduate students, one or two undergraduates) help with the extensive logistics of maintaining a remote field site with no institutional support.
HS: Do you continue to work in this field site? When was the last time you visited Simon’s landing site? In what ways has it changed since the time you worked there for this study? If you were to redraw the community interaction web in Figure 1 today, would it look any different (either because interactions are different today, or because we know more about the system)?
TW: Yes, I was there last week. There have been some large changes recently because of Sea star Wasting Disease that has drastically reduced the starfish population and led to the same sort of changes reported by Bob Paine in his original experiments. The figure wouldn’t change too much by itself, except that the strength of gull predation effects would be diminished. This is because gull populations have been steadily declining over the last three decades, such that gulls are now about half as abundant as when the study started. I do not definitively know the cause, but believe that it is related to changes in the ocean food web. One possible explanation is that bald eagles have been preying on gulls (and other seabirds), their chicks, and their eggs with increasing frequency, because they have had to switch food sources in the face of declining populations of ocean fish such as salmon, along with declines in associated fisheries activities. Some of this decline likely arises from fishing, but we have documented rapid pH decline in the area over the last two decades that likely also plays a role. There is evidence from short-term species dynamics that declining pH is also changing the shore community, making it harder for mussels to maintain their dominance (Wootton et al. 2008 PNAS). Another change is the increasing frequency of elephant seals, which frequent caves that flank each side of Simon’s Landing. They may be responding to protection following the Endangered Species Act, but are also probably responding to shifts in the oceanic food web, where their squid food is thought to be increasing as fin fish decline.
HS: You acknowledge a number of people at the end of your paper. Could you tell us a little more about how you knew them and their contribution to this paper?
TW:
a. B. Mitchell.
Betsy is the wife of Wayne Sousa, visited the island during the study and put in a helping hand in some phase of the experiment (I can’t remember exactly what at this point).
b. C. Pfister
Cathy was a fellow graduate student back then, and now my wife and fellow U Chicago faculty member, who helped put in experiments and provided helpful comments on the study.
c. W. Sousa
Wayne is a marine ecologist at U.C. Berkeley, who visited the island and also gave helpful comments during the publication process, when I was a post-doc at Berkeley.
d. K. Banse
Karl was a biological oceanographer at U. Washington who was on my committee and commented on the original thesis chapter
e. P. Kareiva
Peter was a theoretical ecologist at U. W. at the time who was on my committee and comments on the original study and thesis chapter
Joel is a physiological ecologist/evolutionary biologist, at U. W. at the time, who might have prodded me to think about the Schemske lab’s use of path analysis at some point.
g. G. Orians
Gordon is a famous evolutionary ecologist who was on my committee and commented on the original thesis chapter.
h. R. Paine
My PhD adviser, made lots of comments on paper, during the study, etc.
i. R. Palmer
Posed the original question about birds affecting snails.
j. P. Petraitis
Peter was a reviewer on paper who doesn’t believe in anonymous reviewing (I usually thank all my reviewers). See below.
k. M. Power
Mary was my post-doctoral advisor during publication stage. Made comments on manuscript.
HS: How long did the writing of this paper take? When and where did you do most of the writing?
TW: In the end quite awhile. Probably 3 months to get it written, another year and a half for all the revisions.
HS: Did this paper have a relatively smooth ride through peer-review? Was Ecology the first place this was submitted to?
TW: No, it is one of the hardest of mine to get published. It was only submitted to Ecology. Peter Petraitis (mentioned above) was a reviewer who is not a big fan of path analysis. He provided an extensive review for the first submitted version, provided a second extensive set of new critiques on the revision where I addressed the first round of critiques, and provided a third extensive set of new critiques for the subsequent revision where I addressed the critiques from the second round. Certainly addressing some of the comments were useful (hence the acknowledgement) but I pretty much had to demand that the editor make a decision as continually raising new objections is not a particularly fair approach to peer review.
HS: What kind of attention did this paper receive when it was published? What do you think is the main impact of this paper on research in this topic? Did the approach you proposed (experiments+natural history+path analysis) gain acceptance in the community of researchers studying species interactions?
TW: Certainly some attention, but his is hard to quantify. It has been pretty well cited. It played a part in spurring a group of plant ecologists lead by Jim Grace to advocate more extensively for path analytic approaches and their extension, structural equation modeling. The integration of experiments and path analysis has not been that extensive, however, perhaps because there are still few experimental studies that monitor multiple response variables with enough replication to use the method. Using the method to test alternative causal hypotheses also remains relatively unusual.
HS: What kind of impact did this paper have on your career? In what ways, if any, did it shape the future course of your research?
TW: The story I have heard is that I was nominated for the Mercer Award from the Ecological Society of America (top paper by a young investigator) on the basis of this paper, but it turned out to have been published too recently to qualify, so I received the award for another paper instead. It was also part of the package that let to me receiving the American Society of Naturalists Young Investigator Award. Certainly these awards helped me to be noticed and, in combination with presenting it as part of my job seminars, probably facilitated getting job offers.
I also spent some time thinking about how to use the framework to connect data to dynamical ecological theory, largely because I am a bit cautious about the classical application (using static abundances) being able to disentangle reciprocal interactions (for example, it is hard to see how it could properly assign simultaneously the effect of a prey on its predator (positive expected relationship) with the effect of a predator on its prey (negative expected relationship) without doing an experiment. By instead applying the approach to changes in population size, however, one could estimate parameters in dynamical models, and then use the parameterized model to assess the role of indirect effects in a situation with reciprocal effects (proposed in Wootton 1994 Annual Review of Ecology and Systematics). This has pushed me to collect comprehensive data that documents ecological dynamics on the shore, which I continue to develop to this day.
HS: Today, 23 years after it was published, if you were to re-evaluate the three hypotheses in Figure 3, what would be your conclusion?
TW: The same as in the original paper.
HS: If you were to redo this study today, what would you do differently? Would you still use path analysis? Would you still use a version of the formula you present in the appendix?
TW: I would probably use path analysis, but I would perhaps also have tried to fit a model of species interactions to the experimental dynamics in the data–this has been a direction I have been pursuing as a logical next step to using path analysis, as noted above.
HS: In the paper you say “At present there are few concrete rules that can be applied to determine the best species to manipulate. Natural-history intuition seems to play a large role in determining the species on which experimental community ecologists choose to focus. Thus, formalizing the factors that contribute to natural history intuition represents a potentially useful endeavour. Tentative criteria might include species that exhibit dominance in abundance or biomass, species that have strong patterns of positive or negative association with other species, species that have high rates of resource consumption, or species that serve as resources for many other species.” To what extent has such “formalization of factors that contribute to natural history intuition” happened in the years after this study?
TW: Not as much as I might have hoped, but I currently have a manuscript in the publication pipeline that does this to some extent. It shows that species with high abundance, relatively low variability, and high connectedness to other species tend to be ones with strong experimental evidence of widespread impacts on the rest of the intertidal community, and models of somewhat randomly constructed food webs exhibit the same property. Oddly, I am having at least as much trouble getting this one published as the path analysis paper, but I expect that when it finally is pushed through that it will have a similar impact to the path analysis paper.
HS: In the paper you say “as more possible interactions are included, replication must increase in order to maintain adequate degrees of freedom. Therefore the approach will be more successful if complex communities are organized into sub modules of species (May 1974, Paine 1980) that can be examined somewhat independently of one another; further work is required to determine the reality of such sub modules.” Today, to what extent do we know about the “reality of such sub modules” in communities?
TW: There seems to still be uncertainty about this result. A fair amount of work on the architecture of food web structure has probed this question, and might suggest that overall such sub modules are hard to detect, but more recent work has shown non-random occurrences of sub modules (“motifs”) in food web diagrams that tend to stabilize model food webs. These studies mostly examine patterns of interactions and don’t consider the strengths of interactions, which are an important aspect of defining a sub module in my view. Estimating strengths of interactions remains a daunting problem.
HS: Have you ever read this paper after it was published? If yes, in what context?
TW: Actually, I participated in a seminar/discussion class run by several of my colleagues last fall, and they decided to discuss the paper there, so I had to re-read it and provide some context in that setting recently.
HS: Would you count this paper as a favourite, among all the papers you have written?
TW: I don’t know if I have a favourite, but it certainly is in my top 10.
HS: What would you say to a student who is about to read this paper today? What should he or she take away from this paper written 23 years ago? Would you add any caveats?
TW: First, the problem I was trying to solve is still out there–how do we determine how environmental impacts such as species extinction ramify through complex natural communities? Therefore I think it is a problem still worth pursuing. In the discussion group, I think the students found it most interesting to hear how the paper developed from a simple question that in the end I didn’t have enough statistical power to really answer (the orange snail morph occurred at too low a frequency in the experiment, though the pattern was in the right direction) to something that made a useful contribution to a different area of ecology and evolution. I would encourage students these days to make more of an effort to link their data to more mechanistic models than those represented by path analysis. That being said, I have to admit that path analysis has so far worked more effectively than my proposed modified approach to estimate dynamical models with the data, which raises the question of why this might be. I would also raise the caveats about reciprocal interactions outlined above, and additionally point to one of Petraitis and Dunham’s most important criticisms — the perceived strength of interactions is influenced in part by the degree to which particular species are varied by the environment or an experiment. For example, if I hadn’t manipulated birds, the natural variability of birds is relatively low so a path analysis would likely not pick up their effects very well. What you make of that depends on one’s interest. Is one interested in how important the observed natural variability is, or is one interested in the potential impact an outside perturbation will have on a system?
0 Comments