Tuesday, May 26, 2015

How to respond to reviewers

Previous posts in this "How to" series

4.     How to choose a journal (+ part 2)

Now its time for: How to respond to reviewers

http://evol-eco.blogspot.ca/2011_06_01_archive.html
Students starting in science might imagine that decisions on papers submitted to journals are rather straightforward outcomes such as “accept” or “reject” or seemingly "strong reject". The reality, however, is that most decisions are more on the order of “reconsider after revision” or “reject without prejudice,” or in their more modern forms, “reject with resubmission allowed” or “rejected in present form.” These more ambiguous decisions necessitate a careful revision of your paper in response to reviewer [more properly “referee”] comments, and a cover letter or “response to reviewers” that explains how you dealt with each comment. These revisions and responses are critical to the future of your paper, and the manner in which you implement them will make all the difference between whether your paper ultimately falls into the “accept” or “reject” category. In most of the above categories other than “reject,” your chance of acceptance is actually reasonably high provided you do a good job of the revisions and responses.

Having now published more than 150 papers, having reviewed more than 350 papers, and having been an Associated Editor (AE) for five journals, I have performed and encountered seemingly endless revisions and responses. Some work well and others don’t, and these alternatives typically translate into ultimate accept or reject decisions. This post is an attempt to distill those experiences down to a set of guidelines that can help you to optimize your revisions/responses so as to maximize your chances of acceptance and minimize the number of rounds of review. It is modified from a talk that Jonathan Davies and I give to grad students in Biology at McGill.

Image credit: Nick of http://www.lab-initio.com/.

(By complete coincidence, a related post appeared today at Dynamic Ecology. And here is another recent one at Scientist Sees Squirrel.)

1. The response letter is critical

Most editors and reviewers will make their decision, or at least form a strong initial opinion, based entirely (or mostly) on the response letter, your “response to reviewer/editor comments”. This letter will be the first thing they read and – if they are satisfied with what you say – they might not even re-read the manuscript (MS) itself. Thus, you want to make sure that the editors and reviewers have all of the information optimally organized and explained in one place. Stated another way, the response letter is often just as important (maybe more so) than the changes you make to the MS itself. In preparing your letter, repeat all comments by the Associate Editor and the reviewers and make sure you respond to each immediately below. Repeat not only the negative comments you have to address but also all the positive comments. Repeating the latter is valuable because it can influence the other reviewers: “Hmmmm – the other reviewer quite likes this paper, maybe I am being a bit too harsh.”

As an aside, I often have my students write the response letter even before they modify the paper. This sequence helps to see in advance how their intended actions are likely to play out in the response letter, and doing so helps the students and coauthors to settle on the optimal set of changes. Of course, the letter will need to be modified as changes are made to the MS but it helps to settle the core elements first.

2. You need to convince the reviewers not the editor

Some people attempt to argue to the editor that the reviewer comments are not valid and should be ignored. The thought is that the editor will invalidate those comments and thereby let you proceed without addressing them. In the vast majority of cases, however, the editor will want to get the reviewers to agree to publish the paper and will be unlikely to overrule them (although it does happen). I therefore strongly suggest not trying to argue to the editor that the reviewer comments are invalid. Instead, you need to convince the reviewers, who will then help you convince the editor. In my experience, once you get a reviewer on your side, they will often then help to convince the other reviewer/editor too.

Further to the above, you should not say in your letter to the editor that the reviewer is unqualified or wrong or stupid or sloppy or anything like that. The editor usually selected the reviewers based on who they thought (or who you said) would be good reviewers – often people they know and respect and who could well be their friends. Thus saying the reviewers are unqualified is the same as insulting the judgement of the Editor. Moreover, your response letter – even if written only for the editor – will usually be provided to the reviewers. Thus, something you think you are saying in confidence to the Editor will often make it to the very person whom you are criticizing, which will only further bias them against your paper. Of course, there are exceptions when a reviewer really is personally insulting or overtly biased, in which case you should politely notify the Editor that the reviewer’s comments can be interpreted to be inappropriate. Of course, the other, more modern, strategy is to berate the journal on social media, which can lead them to issue a formal apology. However, I would not advocate this except in extreme cases, such as #addmaleauthorgate.

How editors select reviewers - Grod et al. (2010) Front. Ecol. Evol. 

3. (Re)define the problem

Reviews are often very extensive and different reviewers want different things – often many different things – some of which you can do and some of which you can’t. This can lead to very long and tedious response letters that serve to annoy and alienate the editor and reviewers. In such cases, it helps to define the critical problems at the outset of the response letter (sometimes the editor has helpfully defined them for you.) In essence, you write – just after the editor provides the key points – a short section that explains what you perceive to be the main criticisms, often rephrased in a way that best matches what you intend to do/argue about them. You then provide a short and focused explanation of how you have solved that problem or how, fortunately, the problem does not exist or isn’t too critical. The idea here is get the editor and reviewers focused on just a few key issues and show in a succinct and clear way how you have dealt with them. The implication (and truth) is then that the rest of the comments are really just minor things that didn’t influence the reviewer’s hesitation in accepting your MS. Thus, by showing right up front how you deal with the critical issues that you (or the Editor) think influence acceptance of the paper, you can set them to thinking right away that all will be well.

Types of reviewers - for more details see the awesome page:
http://matt.might.net/articles/peer-fortress/

4. Make actual changes to the MS whenever possible

You can respond to reviewer comments in two ways. First, you can change the MS to accommodate/ameliorate/fix the suggestions or criticisms. Second, you can try to argue your way out of making any changes. Take the first of these two options whenever possible. That is, whenever possible, DO something to the MS that helps to address each comment; and make sure to state that you have done so in your responses. However, it is likely that some reviewer suggestions are either impossible to implement or would actually make the MS worse or would totally change your intended meaning in a way you feel inappropriate. In such cases, you will need to argue your point. However, choosing to argue a specific point also means that you will want to have addressed as many of the other reviewer comments as possible with actual changes to the MS. Making these changes builds good credit with the reviewer and can give you some “free passes” on things you can’t (or don’t really want to) change. In essence, you should be careful to pick your battles, as the more you fight the more likely you are to lose the war. In some of the more difficult cases, you can meet the reviewer halfway by making some partial change (altering a graph, adding a new analysis in the supplementary materials, adding a qualifying statement), and I recommend you do this whenever possible.

As an aside , it is good to do these things even if your paper is rejected and then submitted elsewhere as it is reasonably common to get the same reviewers, who are annoyed at having to say the same thing they said previously and you seemingly ignored. And they really like it if you did what they suggested even if your paper was rejected. This happened to me once. I made many comments on a paper that got rejected from Journal A. I was then asked to review the same paper for Journal B. The author had - even though under no obligation to do so given the switch of journal - implemented essentially all of my suggestions, which were not trivial. I was very impressed and pleased and had many positive things to say about the MS, which as published in Journal B.

5. “I have now made this more clear in the MS”

If the reviewer is clearly wrong about something or if they missed something or misinterpreted something, never say so in as many words. Doing so can seem insulting or condescending (the reviewer missed something that you say is obvious) or it can imply that you think the reviewer is not doing a good job (as indeed they might not be).  Moreover, if the reviewer misunderstood or misinterpreted things, then other readers likely will too, and so you should change it. In essence, a reviewer’s misinterpretation of your study is YOUR fault, not the reviewer’s fault; or at least you should view it that way. In such cases, I first explain the reality and apologize for my mistake (“Fortunately, we did actually do XXXX but it was not sufficiently clear in the MS.”) and then state something like “We have now made this more clear through revisions to the text.” Of course, this means that you do indeed have to make it more clear in the MS – even if you thought it was clear to begin with.

6. Stop whining and just do the new analyses

With ever-increasing statistical sophistication (many would say over-complication), reviewers are likely to recommend some new analysis – no matter how hard you thought out and optimized your analysis in the first place. These new analyses will very rarely change any of your conclusions, and yet the reviewer thought they were important so it is not wise not to ignore them or try to argue them away. By far the simpler solution is to just do the new analysis (or graph or table) and place it in supplementary materials (if you don’t want to change the MS itself). You can then refer to this supplementary analysis in the text with a single sentence referring to the alternative analysis. Although doing the new analysis even when you are confident it won’t change the outcome can seem time-consuming and wasteful, it is even more time-consuming and wasteful to have the reviewer insist again that you do them, thus necessitating another round of revision. Just do the analyses the first time they are requested, which also lets the reviewer know you aren’t trying to weasel your way out of things because you are too busy/lazy or because you have actually done the analyses and found they go against your preferred conclusion. 
Real reviews compiled from my colleagues.

7. Be polite and respectful but not sycophantically so

Sometimes reviewers are insulting (see the above). Sometimes they clearly didn’t see something you had already put in the MS. Sometimes they appear to be complete idiots. Sometimes they are just wrong. However, you should never say any of these things in your response as reviewers will interpret them as being personal critiques. (And, as noted above, the reviewers will likely see things even if you intend them only for the editor.) Thus, no matter how annoying or rude or clueless reviewers are, you can never even hint that such is the case.
Balancing the above point, you should also avoid the temptation to repeatedly say “This is a great comment” or “We thank the reviewer for their comment” and so on. If you say it for every comment, then it means nothing. Instead, save such thank yous for key places, especially where you added an entirely new analysis or, seemingly paradoxically, where you can’t make a change. In this case, you are acknowledging that it is a good point but that you are able to make the change for logistical (or other) reasons.



8. Other procedural points

(a) Don’t paste the exact revised text (or even list the line numbers*) from your paper in the response letter. First, this wastes time as multiple changes are often needed and so the revised text and line numbers keep changing as the revision proceeds. Second, it greatly expands the length of the letter because you still need to explain it. Third, cutting out a single sentence and pasting it into the response letter means that it is out of context (which leads to the next point). Instead, simply explain the change you made and refer to the general part of the paper where you made the change. (Your supervisor or coauthor might have a different opinion here and you should obviously follow theirs in such cases.) *[It has been pointed out to me since writing this that Editors do often like line numbers]

(b) A great temptation is to simply insert a sentence into the MS to address a reviewer comment. Sometimes this works but more often it is out-of-place and out-of-context and awkward – so much so that people reading the paper can clearly see sentences that were added after the fact in response to reviewer comments. Instead, you need to re-read the entire MS (or at least the changed parts) with great care to make sure that any changes you make are seamlessly and effectively integrated into the whole such that they appear to have been there from the beginning. 

Coda

Of course, the above won't always work and, indeed, I have had papers rejected on the second round of review – although rarely. However, I do think the above steps will help you out. In closing, I will leave you with the famous video of Hitler responding to a peer review of his manuscript (which is apparently a lot less funny if you can speak German.)


Monday, May 18, 2015

Cope's Rule: can microevolution explain it?

Body size has been a fascinating subject in science for quite some time. Why were some dinosaurs so big? Why are some organisms so small? What's the optimal size for individuals in a given population? Some of this obsession with size is due to something called Cope's rule. As with all 'rules' it is not a rule, but an idea proposed by Edward Drinker Cope. Funnily enough, he didn't coin the term himself, but he did write about the concept at the end of the 19th century. Basically, Cope's rule is that within a given lineage, body size will increase with time.

Example of Cope's rule where within the lineage that includes today's horses, shows body size getting larger through time. During the Eocene, the ancestors of modern day horses were 10 times smaller.
But before I get into Cope's Rule, I do have to talk about Cope himself. Cope is perhaps most famous (besides having a rule named for him) for being part of the "Bone Wars" where Cope and another paleontologist, Othniel Charles Marsh basically duked it out in person, in literature, and even in nomenclature. For instance, Marsh named one dinosaur Mosasaurus copeanus (divide the species name in half and look at them separately). Cope named one dinosaur Anisonchus cophater and wrote to his assistant Osborn, it's no use looking up the Greek derivation of cophater, because it is not classic in origin. It is derived from the union of two English words, Cope and hater, for I have named it in honor of the number of Cope-haters who surround me.”

Anyways, back to Cope's rule. Cope's rule has been of great interest to researchers, resulting in a large number of papers published in all the glamour journals.

Just a handful of the many papers published about Cope's rule.

Generally, Cope's rule is considered on the macroevolutionary scale. Ok, that's cool, but what could be driving this phenomena? Is there, perhaps, a microevolutionary reason underlying Cope's rule? This is a logical thought as there are lots of individual fitness enhancing reasons to be larger. Larger individuals have been found to have better performance, more dominance, increased mating, and high fecundity. Furthermore, larger individuals can better tolerate short term environmental changes, avoid predators, and extract nutrients. So it would appear that being bigger is better.

And if you look at compilations of selection estimates, this is what has been found. Selection for body size appears to be more often positive, meaning there is increased fitness if the body size is larger.

Figure modified from Kingsolver and Diamond 2011. The red line shows that selection estimates tend to be positive more often, especially when compared with other phenotypic traits. For more information see Kingsolver and Pfennig 2004.

So, if there is positive selection on body size, then this could provide microevolutionary support for explaining the underlying mechanisms of Cope's Rule. Cool! But are populations responding to this positive selection for body size? Populations should be responding to this positive selection by getting larger through time at microevolutionary time scales, and lucky for us, we could actually try to see if this was happening.

These days, there are lots of reviews and studies focusing on contemporary or "rapid" evolution. So much so, that we've assembled a database (that is still continually being added to!) of phenotypic changes. We took studies that had allochronic data: studies of the same population that quantified phenotypes at least twice in time. From, these, we can calculate a rate of change. So, if the rate of change is different from zero, it means the population phenotype is changing through time, and the sign (positive or negative) will tell us the direction of change. In the case of body size, this would mean a positive rate of change indicates the population is getting larger.

So, we calculated rates of change for body size, and just visually, we can see that it seems more of the rates are negative as opposed to positive. This means that populations appear to be getting smaller through time, not larger!

Frequency histogram of Darwin numerators, one of the two types of rates we calculated.
Well, why might this be? Perhaps there is an explanation, such as human disturbances. The database is quite extensive, looking all sorts of populations that have been subject to varying levels of human intrusions. For example, we know that size selective harvesting (think of the cod fishery) can select against larger body size and populations will get smaller. So, we looked at only undisturbed populations, and still, the median rates of change were negative (Darwins = -1087).

Hmm, well, if we run simple, or fancy tests on the data, the results indicate that body size does not appear to be increasing in contemporary populations. If you want to know more about the fancy tests, you can read the article...

So what's happening? If there is positive selection for body size, why are contemporary populations not getting larger? Well, first there are the selection estimates themselves. Selection estimates are an important metric for us evolutionary biologists, but they are not perfect. Selection estimates can be affected by small sample sizes, unmeasured confounding variables, spatiotemporal variation, and imperfect fitness surrogates. Also, there can be a disconnect between selection estimates and actual phenotypic change, and this can happen for various reasons such as countergradient environmental changes, environmental covariance between traits and fitness, and covariance between non heritable traits and fitness. In other words, selection estimates are an important and useful tool, but they might not be the best way to assess microevolutionary trends. Still, our results indicate that body size is not getting larger, and this might be because larger body size does not increase individual fitness as clearly as we thought. Being big means you have to grow faster, and this can lead to increased foraging risk, increased mortality, and structural problems.

Looking at actual phenotypic changes through time in contemporary  populations does not appear to provide a microevolutionary mechanism underlying Cope's rule. However, this isn't really surprising given that there are many differences in micro- and macroevolutionary time scales, and selection can act at different levels (individual vs. population vs. species) and we here focused more on individual level selection. Furthermore, while there is evidence for Cope's rule in the literature, there are also many examples where evidence for Cope's rule have not been found. Perhaps the time has come to retire the rule, and instead, focus on untangling the underlying mechanisms of macroevolutionary trends without names and rules.

Look! We're published!
onlinelibrary.wiley.com/doi/10.1111/evo.12653/abstract


Tuesday, May 12, 2015

Expect the Unexpected, When Measuring Selection in the Wild

[post by Kayla Hardwick, postdoctoral researcher at Reed College]

Imagine a vast expanse of white sand dunes, stretching as far as the eye can see. Now imagine two lizards, one white and one brown, on the white sand. Which do you think would survive better?

Trait divergence between populations can arise via natural selection when particular traits are adaptive in some environments but not others. In the above example, it seems likely that white color would be adaptive in the context of the white sand environment, because the white lizard could more readily blend in with its pale surroundings to avoid predation. But while intuitive explanations for trait divergence often seem straightforward, they can be difficult to test in natural populations. Accurately measuring selection in the wild is challenging, and consequently experimental evidence of the adaptive significance of specific traits is lacking in most systems.

As a graduate student in the early stages of my PhD, I knew (in theory) that measuring selection in natural populations was hard. I had read Endler’s Natural Selection in the Wild, and I was well versed in the common pitfalls of selection studies. But as I read more and more of the published literature on the topic, it seemed to me that measuring selection in the wild typically worked more or less as expected – that most of the time, the authors seemed to confirm their initial predictions without too much trouble. As I embarked on my first large-scale field experiment, with the goal of measuring selection on body color in White Sands lizards, I couldn’t help but think to myself: “how hard can this really be?”

White lizards in White Sands 

White Sands is a gypsum dune field that formed within the Chihuahuan Desert in New Mexico less than 10,000 years ago. The sparsely vegetated gypsum dunes of White Sands contrast strikingly with the surrounding desert scrubland habitat, characterized by dense vegetation and dark brown soil.

White Sands National Monument, New Mexico (photo courtesy of Scott Hardwick)
Three lizard species have colonized White Sands from the surrounding environment: The Lesser Earless Lizard (Holbrookia maculata), the Eastern Fence Lizard (Sceloporus undulatus), and the Little Striped Whiptail (Aspidoscelis inornata). In all three species, lizards in White Sands exhibit blanched dorsal color, while lizards outside of White Sands exhibit darker dorsal color. Blanched color is typically thought to be an adaptation to avoid predation by visually oriented birds that occur at White Sands, such as the Loggerhead Shrike (Lanius ludovicianus) and the Greater Roadrunner (Geococcyx californianus).

Lesser Earless Lizards from White Sands (left) and non-White Sands (right) populations (photo courtesy of Simone Des Roches)
Eastern Fence Lizards from White Sands (left) and non-White Sands (right) populations (photo courtesy of Simone Des Roches)
Little Striped Whiptails from White Sands (left) and non-White Sands (right) populations (photo courtesy of Simone Des Roches)
We have a number of compelling reasons to think that there has been selection on body color in the White Sands habitat. First, color divergence has occurred despite ongoing gene flow between populations on and off White Sands in all three lizard species. In addition, previous research has documented increased genetic divergence between populations at Mc1r, a gene involved in generating the blanched color phenotype. Finally, convergent evolution of blanched color has occurred among disparate taxa at White Sands, including the Apache pocket mouse (Perognathus gypsi), Couch’s Spadefoot Toad (Scaphiopus couchii), and the White Sands camel cricket (Daihinoides hastiferum).

The White Sands camel cricket (photo courtesy of Scott Hardwick)
Given these lines of evidence, the most intuitive explanation for white lizards in White Sands is natural selection for cryptic coloration. We thus set out to answer a seemingly simple question: does blanched color make White Sands lizards less susceptible to bird predation?

Painting lizards, for science!

To study the effect of color on survival, we conducted an enclosure experiment with substrate-matched and substrate-mismatched Earless Lizards in the White Sands habitat.

Including both matched and mismatched lizards in our study required some creativity – in the “arts and crafts” sense – as we used human temporary-tattoo paint to manipulate the body color of White Sands Earless Lizards. During graduate school, students seem to acquire at least one skill that is totally bizarre and has no practical applications in real life. For me, lizard painting is that skill. We painted lizards to be either substrate-matched or substrate-mismatched, mirroring the colors of the naturally occurring Earless Lizard color morphs as closely as possible.

Painted, substrate-matched and substrate mismatched Earless Lizards (top panel), and naturally occurring Earless Lizard color morphs (bottom panel)
We built enclosures within the White Sands habitat using steel flashing and rebar, which turned out to be a uniquely challenging experience. Most locations in White Sands are not accessible by car, which meant we spent a good deal of time carrying heavy equipment up and down sand dunes. The thing about sand dunes is, they tend to collapse beneath your feet as you walk. Thus it was not uncommon to find yourself, with a 50lb roll of flashing under your arm, taking dozens of steps up the side of a sand dune without actually going anywhere. This particularly frustrating phenomenon came to be known among my field assistants as “Nature’s Stairmaster.”

But despite these (and other) difficulties, we ultimately managed to build a total of four enclosures in White Sands, each measuring 10 meters by 10 meters. We divided each enclosure in half and left one half uncovered to allow avian predators to enter and exit freely (the “open” treatment). We covered the other half of each enclosure with chicken wire to exclude predators (the “closed” treatment). We then released painted, substrate-matched and substrate-mismatched individuals into the enclosures. We recaptured surviving lizards after two weeks.

The “open” enclosure treatment allowed bird predators to enter and exit freely
The “closed” enclosure treatment was covered with chicken wire to exclude predators
So, what do you think happened when we put substrate-matched and substrate-mismatched lizards into enclosures, and left them vulnerable to predation? I’ll give you a hint: it’s definitely not what you’re expecting!

Result #1: Lizards in White Sands get eaten by birds

Okay, so maybe you were expecting this particular result. We found that lizard survival differed significantly between the open and closed enclosure treatments. 100% of lizards in the closed treatment survived trials, compared with just over a third of lizards surviving in open treatment replicates. This indicates that lizards in our trials experienced predation. Birds were most likely responsible for lizard predation in our study; we observed Loggerhead Shrikes perched near enclosures on a number of occasions, and found Greater Roadrunner tracks inside enclosures multiple times throughout the trials.

Result #2: Lizards get eaten, no matter what color they are

Despite high rates of predation within the open enclosure treatment, we did not detect selection on color in our study. In other words, there was not a significant difference in survival between substrate-matched and substrate-mismatched lizards. This result is unexpected, given that previous studies offer strong evidence that blanched color is adaptive in White Sands. In a system amenable to experimental manipulation where divergence is literally in black and white, why did we fail to detect selection on this putatively adaptive trait?

Maybe there isn’t selection on color at White Sands

It is possible that there is not currently selection on body color in White Sands lizards. Selection studies can be complicated by differences between historical and contemporary selection pressures, where traits that originate as adaptations can remain widespread in a population even after shifts in the ecology of the system render those traits unnecessary for survival. The distributions of key predators in the White Sands system, such as the Loggerhead Shrike, have shifted dramatically in recent decades. It is thus possible that blanched color at White Sands, even if it was historically important for survival, could represent a “ghost of selection past” today.

Then again, maybe there is selection, and we just didn’t detect it

Previous research has given us reason to believe that selection on color is ongoing in White Sands lizards. If there is indeed contemporary selection, are there any compelling explanations for why we might not have detected it in our study?

Variable selection can complicate studies of the adaptive significance of specific traits. Previous research indicates that selection pressures on particular phenotypic traits can vary between sexes and among life stages, as well as over space and time. When selection is variable, it can be difficult to discern the adaptive significance of specific traits because the outcome of a study may depend on the study’s scope. For example, if selection varies among different life stages, measurements of the strength and direction of selection may be contingent on the number of individuals of each life stage included in a study.

In our experiment, there was a hint that selection on color varies between males and females. Specifically, substrate-matched males tended to survive better than substrate-mismatched males, while in females the opposite was true. This could indicate that substrate matching is more important for survival for males than females in Earless Lizards, as males are typically more behaviorally conspicuous than females in iguanid lizards.

Our results also indicate that selection on color is variable over space and time. Both year and enclosure location affected survivorship in our study, where survival was higher in some enclosures than others, and this pattern changed from year to year. Spatial and temporal fluctuations in predation in our study could have been related to the physical proximity of enclosures to suitable perches/nest sites, seasonal changes in predator behavior, and any number of other biotic/abiotic aspects of the environment.

Variable selection can also interact with issues related to statistical power, increasing the number of individuals required to discern the adaptive value of specific traits. For instance, the sample size required to detect a significant effect of color would have been substantially smaller had we focused exclusively on males (thereby avoiding the effect of sex-specific variations in selection), and smaller still if selection against substrate-mismatched individuals had remained consistent throughout replicates (thereby avoiding the effects of spatial and temporal variation). White Sands is a small, isolated population, and variable selection thus presents a significant obstacle for future studies of selection in this system.

Finally, it can be difficult to replicate natural conditions in an experimental context, and deviations from natural conditions could have affected the results of our study. In particular, the density of lizards within our enclosures was substantially higher than that at which Earless Lizards naturally occur in White Sands. It is possible that we observed a "buffet effect" in our study, where predators were initially attracted to the enclosures by substrate-mismatched lizards, but subsequently proceeded to consume lizards of both paint treatments. It is also possible that the high density of conspecifics caused lizards to behave more conspicuously that they otherwise typically would, further attracting the attention of predators to enclosures.

Painted lizards in an enclosure trial (photo courtesy of Susan Kologi)

So, what’s the verdict?

There are a number of compelling potential explanations for the results of our study. The bottom line is, we don’t yet know exactly what’s going on in terms of selection on color in White Sands. Future research will focus on figuring out why substrate matching might affect males and females differently, identifying factors that contribute to spatial and temporal variability in predation, and looking into the importance of selection on correlated traits (including the thermoregulatory effects of blanched color). Read our recently published manuscript for a more detailed discussion of our findings, and check out the Rosenblum Lab website to learn more about the evolution and ecology of White Sands lizards.

Challenges such as those discussed above are quite common during studies of selection in the wild, and their causes and consequences can be difficult to predict. And yet, many published selection studies seem to minimize (or completely ignore) the effects of such challenges, perhaps because those that freely discuss possible methodological shortcomings are less likely to be accepted for publication. This potential publication bias is important to acknowledge, because researchers do not yet have a comprehensive understanding of how different challenges manifest themselves in terms of the results of a study. Publishing the results of rigorous experiments, even when they do not turn out exactly as planned, is absolutely crucial for facilitating future research in the field.

Measuring selection in the wild is a challenging, complex, and incredibly important endeavor. The results of our study indicate that, despite the difficulties detailed above, researchers often learn fascinating things about selection when experiments have unexpected outcomes.

Thursday, May 7, 2015

The Adaptive Radiation of Darwin's Finches: old, new, and personal perspectives


[Al Uy asked to me to write a pseudo-popular piece about the adaptive radiation of Darwin's finches for a forthcoming "incipientspecies.org" website. I decided to use this blog to present a first (and as yet overlong and too technical) draft of my contribution.]



When I went home for Christmas in 1995, I was an aspiring salmon biologist doing my Master’s research at the University of Washington, Seattle. By the time I returned to Seattle two weeks later, I was an aspiring evolutionary biologist. Now, 20 years later, I suppose I am an established evolutionary biologist, although I am still aspiring! I have worked on stickleback in British Columbia, guppies in Trinidad, and – of most significance to the present story – Darwin’s finches in Galapagos. I doubt I would have worked in any of these systems had it not been for one gift that fateful Christmas – a book.

The book was The Beak of the Finch by Jonathan Weiner, something my mother had seen in a book store and thought I might like. For the next few days, I sat over the heater vent entombed in a blanket with only a pane of glass between me and snowy −25°C Edmonton, losing myself in the tale of two Princeton University evolutionary biologists, Peter and Rosemary Grant, and their quest to understand how evolution works through detailed long-term studies of Darwin’s finches. Most amazing to me was the description of how this work had demonstrated evolution occurring on very short time scales, sometimes only a single generation. I had never imagined that it might be possible to watch evolution take place in almost real time – yet they had done it. From that moment onward, I wanted to do the same thing and, within a few days of returning to Seattle, I went to the library and photocopied every paper by Peter and Rosemary. Slowly, the story emerged.




An image of the Galápagos Islands and their topography. Image from Wikimedia Commons. 

The Galapagos Islands were formed by magma welling up from a hotspot on the ocean floor to build underwater mountains, some of which broke the surface and became islands. When the first island formed, more than 8 million years ago, it had no terrestrial life given its 900+ km separation from the mainland. With time, however, various plants and animals either flew, drifted, or were carried to the islands. One of those colonists, arriving approximately 1.5 million years ago, was a bird – presumably a flock of them – that probably looked something like a modern-day grassquit.

The Darwin’s finch ancestor may have looked this Black-faced Grassquit (Tiaris bicolor). Photo by C.J. Sharp on Wikimedia Commons.
These colonizing proto-finches arrived in an ecosystem that had been assembled from the relatively few long-distance migrants that had reached the islands, some of which had then diversified into multiple species on the islands. This initial finchless ecosystem had a number of different potential food types (insects, fruits, seeds, and leaves of various sorts), but very few – if any – other birds to eat them. These first colonists were thus confronted with a land of “ecological opportunity” filled with a number of “empty niches” that might be filled by finches.

The colonizing finches increased in abundance and spread across the various Galapagos islands. As they did so, they encountered different conditions. Some islands were very low and dry. Some islands were high and wet. Some had many insects, some had few. Some had certain types of plants, some had other types. Each of these different sets of conditions meant that a different way of feeding would be optimal at different locations. These different ways of feeding had a classic evolutionary target – the beak of the finch.

The diversity of pliers. Images from http://coilku.blogspot.ca/ 
Bird beaks have been likened to pliers and, like pliers, different shapes and sizes are best suited for different tasks. The thin, pointed beak of a warbler is well-suited for gleaning insects, the long beak of a honeycreeper for probing flowers for nectar, the chisel-like beak of a woodpecker for tearing apart wood, and the robust rounded beak of a finch for cracking seeds.

The diversity of bird beaks and their functions. Image from L. Shyamal on Wikimedia
Thus, as the proto-finches spread across Galapagos and encountered different conditions, the different populations began to experience natural selection for different beak morphologies. Over perhaps a relatively short period of time, adaptation drove those different populations toward those different beaks, yielding a small pointed beak for insect eating (eventually to be called “warbler finches”), a long bill for nectar feeding (eventually to be called the “cactus finch”), a chisel-like bill for tearing apart wood (eventually to be called the “woodpecker finch”), a robust beak for cracking seeds (eventually to be called the “ground finches”), and so on. Yet, if it is geographic variation in food types that drives this “adaptive radiation” of finches, how does one end up with multiple species at any given location – as is currently the case in Galapagos?

One representation of the Darwin's finch radiation (Grant 1986: Ecology and Evolution of Darwin's finches)
Given that the proto-finches spread to the diverse locations in Galapagos in the first place, it seems just as plausible that newly-evolving species could similarly spread to different locations, some of which would already host locally-evolved species. If the end result of this “secondary contact” is to be a multi-species finch community, two requirements must be met. First, the invading species has to have a different diet than the resident species – otherwise one species will simply out-compete and thereby exclude the other species (“competitive exclusion”). Second, the invading and resident species can’t interbreed too much – otherwise they will simply fuse together into a single species. Fortunately, these two requirements often seem to be met for Darwin’s finches in Galapagos.

Crude depiction of the distribution of different finch species on the different islands. From A Field Guide to the Birds of Galapagos by Michael Harris. Collins. 

First, the previously described situation in which populations in different places show adaptation to different food types means that species coming into secondary contact have already specialized on somewhat different food types, thus reducing competition. This initial divergence can increase following secondary contact due to selection against individuals that have traits/diets/behaviors most similar to the other species, and that therefore experience the highest competition. The resulting process of “character displacement” will then further reduce competition and promote species coexistence.

David Lack's classic demonstration of character displacement. Image from Ricklefs' (1996) Economy of Nature
Second, the traits (beak size and shape) that undergo adaptive divergence influence mate choice so as to reduce interbreeding. In particular, beak size is strongly correlated with the types of songs that males can sing: for instance, large-beaked individuals cannot sing rapid and complex songs. Thus, adaptation to different food types should cause divergence in songs as an incidental byproduct. Moreover, offspring tend to “imprint” on the songs of their fathers and, at maturity, male offspring sing those songs and female offspring prefer similar songs. As a result, birds that have evolved different beaks automatically show reduced inter-breeding – making beaks an outstanding candidate for the so-called “magic traits” of speciation. Moreover, any successful interbreeding between species (which does occur reasonably often) produces offspring with intermediate beak sizes that have low survival rates because they are not well-suited to the diets of either parental species.

Different finches sing songs with different vocal properties. From Podos (2001 - Nature).

This rough thumbnail sketch provides a crude summary of the process of adaptive radiation in Darwin’s finches as it was understood at the end of the 20th century.



Although I now, at the start of 1996, wanted to be an evolutionary biologist, it never occurred to me that I might actually work on Darwin’s finches – they were simply too far away in space and too much a place of my imagination rather than reality. Instead, I turned my attention to studying how evolution works in fishes, where I already had some experience; but serendipity intervened. In 1998, I started a postdoctoral position at the University of British Columbia (Vancouver, Canada) studying the evolutionary biology of threespine stickleback. A short time later, I saw an ad for the “Darwin Postdoctoral Fellowship” at the University of Massachusetts in Amherst, Massachusetts. “Wow, what a cool name for a fellowship,” I thought, “maybe I should apply.”

One of my stickleback experiment field sites on Vancouver Island.
Less than a year later, I was ensconced in the Department of Organismic and Evolutionary Biology at UMASS in Amherst. My project was a logical extension of my PhD work and focused on natural selection acting on introduced Atlantic salmon in a restoration project for the Connecticut River. While at UMASS, I became friends with a new faculty member across the hall, Jeff Podos, who had started working on Darwin’s finches and had just published a paper in Nature about the vocal constraints faced by birds with different beak sizes. Soon afterward, Jeff received an NSF “Early Career” fellowship that enabled him to – pretty much – do whatever he wanted for research. So Jeff started assembling a team for new Darwin’s finch work and asked if I wanted to come along. Twist my arm.

Some of the Darwin’s finches I first encountered. Clockwise from upper left: medium ground finch (Geospiza fuliginosa), medium ground finch (Geospiza fortis), large ground finch (Geospiza magnirostris), cactus finch (Geospiza scandens), small tree finch (Camarhynchus parvulus), vegetarian finch (Platyspiza crassirostris), and woodpecker finch (Camarhynchus pallida).
In 2002, after all those years of reading and thinking about Darwin’s finches, there I was actually in Galapagos looking at Darwin’s finches hopping about on the ground making, as David Lack had said, “dull unmusical noises.” My goal that first year was simply to learn as much as I could about the finches and to think broadly and creatively about various projects that I might do in collaboration with Jeff and his team. One of the highlights that first year – perhaps even a rite of passage – was the afternoon I spent walking around in the field with Peter and Rosemary. Now, 15 years later, a number of projects have come to fruition, some of which have modified the basic tale of the adaptive radiation of Darwin’s finches as described above.



1. Darwin’s finches are “imperfect generalists”

As described earlier, a critical mechanism by which two young Darwin’s finch species coexist when they come into secondary contact is through adaptation to different resources. The assembly of a community of finches thus depends critically on the extent to which different species partition their resources. One of our first goals was to understand how this partitioning took place, so – starting in 2003 – we began what would become a long program of simply walking around our field sites, finding birds, identifying them (through binoculars) to species, and determining on what they were feeding. This task has been greatly facilitated by the fact that Darwin’s finches are very tame. In the early stages of this work, we were quite surprised to see that, contrary to our initial naïve expectations, most of the species seemed to be feeding on pretty much the same things. Where was this niche partitioning that was supposedly so critical to the adaptive radiation?


Those first years were very wet, with lots of plant reproduction, lots of seeds, and lots of insects. But then a major drought occurred and, for several years, plant reproduction was minimal and so seed and insect abundances declined dramatically. Fortunately, we had continued to record what the finches were eating throughout this period. During these drought years, we found that the different species increasingly diverged to use different resources – and niche overlap decreased accordingly. Thus, with 5 years of feeding observation data spanning wet and dry years, we were able to conclude: These results together suggest that the ground finches are ‘imperfect generalists’ that use overlapping resources under benign conditions (in space or time), but then retreat to resources for which they are best adapted during periods of food limitation. These conditions likely promote local and regional coexistence (De Leon et al. 2014). This finding that niche overlap decreased in years when little rain fell fit well with earlier observations that niche overlap decreased during the dry (as opposed to wet) seasons within a year.



2. The adaptive radiation is ongoing

During that early walk in 2002 with Peter and Rosemary, I asked them what they thought would be one of the most interesting questions to investigate on Santa Cruz, where we were planning to work. They suggested trying to understand the causes and consequences of the hyper-variable population of medium ground finches, Geospiza fortis. It turns out that this species is more variable on Santa Cruz than anywhere else – indeed, they are so variable that a paper by Hugh Ford in 1973 argued they were undergoing sympatric speciation. Just the next year, team member Anthony Herrel came into our dorm room to show us some data from the birds we had captured that year. The histogram of beak sizes was bimodal – just like Hugh Ford had reported. This result was inspiring because it suggested that the population might be in the midst of splitting into separate species, a rare event that would enable us to formally test the mechanisms thought to promote the adaptive radiation of Darwin’s finches. (It is difficult to test such mechanisms when species are already well established.)

We first formally confirmed that the G. fortis population under study (at El Garrapatero on Santa Cruz) was indeed bimodal (Hendry et al. 2006). We then used this bimodality to test a series of predictions stemming from the theory of adaptive radiation – and the results confirmed expectations. (1) The two morphs had different diets, with the large morph generally eating larger/harder seeds (De Leon et al. 2011, 2012). (2) The two morphs differed in their feeding performance, with the larger morph that fed on larger/harder seeds having higher bite forces (Herrel et al. 2005, 2009). (3) Males of the two morphs sing different songs and respond differently to each other’s songs (Huber et al. 2006; Podos 2010). (4) The two morphs mate assortatively: small females with small males and large females with large males (Huber et al. 2007). (5) The two morphs experienced “disruptive selection” in that intermediate birds survived at lower rates (Hendry et al. 2009). (6) Gene flow was somewhat limited between the morphs in that they showed some (albeit minor) differences at neutral genetic markers (De Leon et al. 2010). All of these findings supported the general model for adaptive radiation in Darwin’s finches: different diets leads to song divergence leads to different traits leads to reproductive isolation.

Clockwise from top left: The large and small beak morphs, the demonstration of bimodality, assortative mating, and disruptive selection

3. Human influences on adaptive radiation

Given a population of finches seemingly in the midst of splitting into separate species, none of the above results were surprising – yet we did have a surprise coming. The bimodal population of finches described above was not the same population previously described by Hugh Ford. For the latter population, found at “Academy Bay” immediately adjacent to the main tourist town of Puerto Ayorra, we couldn’t find evidence for bimodality in any of our new samples. Even though the population was still quite variable, it just didn’t show the dip in the frequency distribution of beak size that Ford had reported and that we were finding at El Garrapatero, just 7 km away. It seemed that Ford’s population had lost its bimodality between then and now, and we became curious as to just when that collapse had occurred. At this point, we rounded up all of the previous data available for G. fortis from Santa Cruz. Peter and Rosemary and their collaborators provided much of it, including data from David Snow’s collections in 1963–1964. The most fun for me, however, was to find Hugh Ford’s contact information online and send him an email:

In 1973, you published a wonderful paper arguing for bimodality in beak size distributions in G. fortis at Academy Bay. I have recently compiled morphological data from Peter Grant and other investigators from a variety of sites on Santa Cruz island. I am now testing for spatial and temporal variation in the extent of bimodality. This work would be greatly aided if you happened to have the raw data from your 1973 paper. … I hope that you have these data and would be willing to send them to me. It would be much appreciated.

To which he replied:

Amazingly I have found the data in my room. It is in old notebooks - no Excel spreadsheets in those days! I can enter the data into a spreadsheet but not right away. … Do you want date of banding - all August to October 1968 I think?

And so, not much later, I had the original data in hand, which was particularly exciting given that it was collected in the year I was born!

Compiling all of these data made the picture clear – bimodality had decreased substantially in Academy Bay G. fortis some time soon after Hugh Ford’s study (Hendry et al. 2006). Noticing that these decreases coincided closely with an exponential increase in human population density, we argued that humans were altering the food resources that had been so fundamental to shaping finch diversity. By subsequently comparing contemporary G. fortis from Academy Bay and El Garrapatero, we provided support for this hypothesis by showing that associations between beak size, diet, bite force, and genetic variation were all now much weaker at Academy Bay than El Garrapatero. Humans, it seems, can alter the “adaptive landscape” for finches and thereby reverse the process of adaptive radiation (De Leon et al. 2011). Whether or not such effects can extend to the more discrete species of Darwin’s finches, which do still hybridize, remains to be seen.

Top left: Academy Bay. Bottom left: finches eating rice provided by humans. Right: bimodality at Academy Bay the year I was born and at El Garrapatero when we started our work, but not at Academy Bay in the same year (2004).





From the inspiration initially provided by that one book my mother gave me back in 1995, my career has followed quite an arc, to the point where I am now contributing knowledge to our understanding of the adaptive radiation of Darwin’s finches. Sometime in the mid-2000s, I received a phone call from Jonathan Weiner, the author of that fateful book. He wanted to talk about some related work I had done on fishes. What fun it was to be able to tell him how influential his book had been to my own career path. But where to now?

The next frontier for Darwin’s finches is genomic work. Just this year, an initial study was published exploring genomic variation among the various finch species – and we have begun our own work on the topic. Over the next decade or so it seems likely that we will have answers to many questions that research on adaptive radiation has long pondered: how many genomic regions are involved, how big are the effects of particular genes, which specific genes are involved, and are the genes involved now (at the tips of the ongoing radiation) different from those that were involved earlier (at the deeper splits of the radiation)?


Another critical frontier is to examine how finches influence the evolution of plant traits and assembly of plant communities – work we are now starting with Marc Johnson, Nancy Emery, and Sofia Carvajal. Much work remains to be done, and I am curious to find out just how Darwin’s finches will continue reshaping my life and career.



Friday, May 1, 2015

Lessons from The Stickleback

 Lessons from The Stickleback: an undergraduate’s first quest into research and publication in evolutionary ecology

I found myself enrolled in an honours biology class at the University of British Columbia in my third year. Everyone else in the class had already found a lab to work in. I, on the other hand, hadn’t a clue about research, other than the fact that I had to start some soon. I had very little idea about how I was to find a lab to work in, let alone a project*. Thus I began my quest for potential advisors. I liked “ecology.” I liked “evolution.” So I searched those (painfully general) terms through the UBC Zoology faculty websites. Eventually I found myself at the door of Dolph Schluter's office. Dolph literally wrote the book on adaptation and ecology. But I was a lowly undergrad, and I (thankfully) knew nothing about fame in academia.

Dolph described a project that examined the conventional ecology–evolution relationship from an entirely new perspective. Well, the opposite perspective anyway. I knew that evolution could occur via species adapting to their ecological surroundings – that is, how ecology affects evolution. The proposed project, however, looked at the ecology–evolution relationship from the other direction – how do specialization and diversification that result from evolution then feed back onto the species’ ecological surroundings? To examine these questions, the project would employ Dolph’s primary study system, the threespine stickleback.


My illustration of limnetic and benthic stickleback, which did not make it onto the cover of any journal.

I knew about character displacement and the ecological niche, and I daresay I even know about these concepts in threespine stickleback. Namely, several freshwater lakes in coastal British Columbia contain two “ecotypes” – not quite species by most definitions, but populations with specialized feeding morphology and behaviours suited to specific ecological niches. The “limnetic” stickleback, with several small gill-rakers, a narrow gape, and slender body, feeds on tiny zooplankton in open water. The “benthic” stickleback, with fewer large gill rakers, a larger gape, and deep body, consumes larger macrinvertebrates from the lake bottom (Schluter 1994). Legend (read, plentiful research) has it that the two ecotypes most likely experienced ecological divergence through two colonization events (from the ocean via repeated sea level rises) followed by character displacement: the first colonist evolved into the benthic ecotype, and the second became displaced to the limnetic niche (Schluter and McPhail 1992). Although the limnetic-benthic stickleback divergence has occurred in parallel in several freshwater lakes, most lakes contain one, solitary, generalist type, which has intermediate morphology between the two ecotypes. In some cases, such as on Texada Island, “species pair” and “solitary” population lakes are within a mile (1.6 kilometers) of each other.




If you are ever my advisor, you will undoubtedly receive artwork: illustrations for Luke Harmon and Dolph Schluter, respectively, of a “species pair” stickleback community (top) and an unfortunate looking solitary generalist (bottom).

We would test a few different hypotheses regarding the effect of evolution (stickleback diversification and specialization) on ecological parameters. Because the overall ecological effects of stickleback presence were known (Bell et al. 2003), and stickleback population size probably also changed with colonization and diversification, we also decided to include treatments that varied in stickleback density. In fact, the comparison between the effects of differences in stickleback evolution and stickleback density on ecological surroundings became the main focus of this particular project.

The experimental set-up and my role in the project were straightforward. I was introduced to Luke Harmon (who later became one of my PhD advisors) and Jon Shurin (who co-advised my honours thesis). Together we created miniature... well, medium-sized ecosystems (mesocosms) of six different treatments: no fish, generalist fish only, limnetic fish only, benthic fish only, limnetic and benthic fish together, and finally a double-density treatment of both limnetic and benthic fish. My main job was to sample and measure different ecological response variables indicative of the abundance of organisms at different levels in the food chain (periphyton, phytoplankton, and zooplankton) and primary productivity (dissolved oxygen). Collecting fish on Texada Island, staying up late and waking up early to measure dissolved oxygen in each of the forty tanks, consumed by counting zooplankton species under a dissecting microscope, and – most dreaded – filtering chlorophyll from periphyton and phytoplankton samples. Such was my first summer of research. I was finally a real scientist, studying something that had basically never been studied before.

I learned that science, especially when it’s outside, is messy. We had a number of fish deaths, and every time I found a little stickleback floating belly-up, I had to rush back to the aquariums on campus and find an individual matching it as closely in size and weight as possible. One tank killed all its fish and had to be abandoned entirely. One tank hatched long-toed salamanders (which I dutifully helped disperse elsewhere). Several tanks became nurseries to baby stickleback (after our last sampling date, thank Gaia). And the tanks became more and more difficult to maneuver among during sampling – we had cleared an area of bushes and narrow trees to set up the tanks, but come the springtime in Vancouver, the blackberries were making a ferocious comeback.


Mesocosms in the early early morning... slowly being overtaken by blackberry bushes

Alas, results emerged. Both stickleback evolution (diversification and specialization) and ecology (density) affected ecological variables in some way. When compared to the generalist treatment, the limnetic treatment, the benthic treatment, and the limnetic–benthic treatment differed in zooplankton community composition (that is, the type and abundance of different zooplankton). Productivity – both as indicated by dissolved oxygen and chlorophyll from periphyton and phytoplankton – differed significantly between density treatments, as did abundance of small zooplankton (rotifers). Specifically, productivity and abundance of small zooplankton increased with the presence of, and with increasing density of, stickleback (an expected result, considering the predictions of the trophic cascade hypothesis).


Effect of different treatments on primary productivity: C) phytoplankton and D) periphyton chlorophyll-a, and E) dissolved oxygen concentration.

Illustration for Jon Shurin showing an awkwardly-purple-eyed stickle eying a diaphanosoma.

Interesting results, we all thought. Let’s do it again, said Dolph. Let’s narrow it down and more specifically ask about those evolutionary effects. Let’s change the experimental set-up and abandon the no-fish and double-density treatments, use only female fish so no one breeds, move the whole experiment to a new, less cluttered site. Let’s take on another collaborator (Blake Matthews) who could help us measure other ecological variables like dissolved nutrients, the light environment, and benthic macroinvertebrates. And finally, let’s measure how these ecological variables change over time.


Pristine mesocosms for the Harmon et al. (2009) project.

The results, which I will not discuss here, were published in Nature (Harmon et al. 2009), and saw a lot of publicity. I was, and still am, infinitely grateful to have been included on this project and publication. At the time of the Nature paper, I was starting my PhD on another system. Unfortunately, I was facing delays with publishing the results of the first project (which was a considerable challenge after the Nature paper’s release). Turns out the final lesson the stickleback had for me was how to distinguish my paper (on a project that we performed first) from a substantially more prominent publication (on a project that we performed second). Reading the publications now, in light of the above, I hope you will see how different the projects really were. In 2013 I published the first project in the very popular open access journal PLOS One.

In hindsight, it is clear to me how fitting it is that the two projects were published in such different ways. The avenues of publication, I think, were not necessarily reflective of the importance of the results, but rather, were a consequence of publication order and different approaches to disseminating the work. The Nature paper was not open-access, but has become widely read. The PLOS One paper is open access, but is (I must admit) probably relatively hidden among the thousands of papers published in that journal every year. However, I can say now that publishing in PLOS One allowed me to preserve the integrity and, okay, the soul of my undergraduate research. I got to keep all my Figures. My Methods are in a full-size font right after the Introduction, where, after so much of my blood (recall, blackberries), sweat, and tears, they deserve to be. This paper is no shadow of Harmon et al. 2009. If anything, by looking at both ecological and evolutionary effects of stickleback, it provides a solid context for studies like Harmon et al. (2009), which examine only the latter.

I can’t think of a better way to be introduced to the fields of ecology and evolution than with projects that look at their dynamic relationship from a new point of view. In my mind, there is no question that ecological and evolutionary processes feed back on one another. The strength of these interactions, I imagine, depends largely on the system examined, and I look forward to reading about new projects in which the mechanisms behind the eco–evo relationship are explored.


* Every time I tell my ~30 first-year students to get involved in research as early as possible, I feel a twinge of irony as I think back to my immense reluctance to do so myself.


References

Bell T, Neill WE, Schluter D. (2003). The effect of temporal scale on the outcome of trophic cascade experiments. Oecologia 134: 578–586.

Harmon LJ, Matthews B, Des Roches S, Chase JM, Shurin JB, et al. (2009). Evolutionary diversification in stickleback affects ecosystem functioning. Nature 458: 1167–1170.

Schluter D, McPhail JD. (1992). Ecological character displacement and speciation in sticklebacks. American Naturalist 140: 85–108.

Schluter D. (1994). Experimental evidence that competition promotes divergence in adaptive radiation. Science 266: 798–801.