Don’t ask “When is it coevolution?” — ask “How is it coevolution?”

A rough-skinned newt, Taricha granulosa, somewhere on the Marin headlands of California. Rough-skinned newts’ geographically varying arms race with predatory garter snakes has made them a classic case study for “geographic mosaic” coevolution. (Flickr, matt “smooth tooth” knoth)

Ask me to pick a single word that describes what I study, and I’ll typically say “coevolution.” This is probably true of most evolutionary biologists who study interactions between species — plants and pollinators, hosts and symbionts, predators and prey, et cetera and so on. I can also rattle off a definition drilled into my memory by repeated exposure in graduate school, and then just repetition over a half-dozen semesters teaching evolutionary biology: Coevolution is reciprocal adaptation of interacting species. We usually understand that to mean coevolution is specific, that the interacting species are interacting one-on-one; and that the adaptation is more or less simultaneous, that one species adapts to adaptive changes in the other as those changes occur, and vice-versa. So really, coevolution is specific, simultaneous, reciprocal adaptation of interacting species.

Like a chromosome unspooling into megabases of DNA sequence, that definition unpacks into a huge research effort. Evaluating whether or not adaptation has occurred (or is occurring) requires identifying genes or genetically determined traits involved in a species interaction, then testing for evidence of natural selection on those genes or traits — either relationships between trait values and growth, survival, or reproductive success, or else population genetic patterns consistent with a history of selection. That’s a lot of work when you’re studying the adaptation of one species. Make it two, and you’re beyond what may be possible in a single doctoral dissertation, or even a collaborative grant proposal.

Lately, though, I’ve started to wonder: what if we don’t actually need to do all that work?

To be clear, research to document and measure reciprocal selection is unquestionably valuable! Evolutionary biology is about understanding what forces of evolution shaped biological diversity, and every study that quantifies natural selection contributes to that understanding. What I have come to wonder, though, is whether we need to demonstrate coevolution in the strict sense — specific, simultaneous, reciprocal adaptation of interacting species —  to answer the questions we have about the evolutionary past and future of ecological communities.

This post is an effort in talking through the reasoning behind that question, and my first attempt at an answer to it. I hope it’ll be a preview of a more formal presentation (journal editors, drop me a line) and that readers will let me know where they see me missing important points of fact, history, or logic. (Or, indeed, whether I’m rediscovering a question someone else has already raised and dispatched.)

The earliest published usage of the term coevolution in the sense we use today doesn’t provide an explicit definition, and doesn’t even use the modern spelling, but it has a crystal clear reason to be interested in specific, simultaneous, reciprocal adaptation — because it relates to crop productivity. Charles Mode’s 1958 paper “A mathematical model for the co-evolution of obligate parasites and their hosts” models the inverse-frequency-dependent host-parasite coevolution implied by then-recent work on the genetics underlying infectivity and resistance for flax rust and flax and mildew and barley. The significance of this process, as Mode says, is that it maintains diversity at the interacting genetic loci of the parasite and the host over very long periods of time, going back before the domestication of crop species. Mode’s intended meaning for co-evolution is clear from context, and the model itself is the definition of that process.

Leaves of flax infected by flax rust. (Wikimedia Commons, Carl Davies/CSIRO)

The term and the process didn’t become synonymous for a while, though. In a 1961 paper discussing the “genetic feed-back mechanism” created by the interaction of “herbivore and plant, parasite and host, and predator and prey”, David Pimentel never uses the word coevolution, hyphenated or otherwise — he simply refers to the “evolution” of the species interactions he discusses. (He doesn’t even cite Mode!) Then in 1964, Paul Ehrlich and Peter Raven’s foundational “study in coevolution” examined congruences between major clades of butterflies and the plants they use as larval food sources, and proposed a process to explain it that need not be specific, simultaneous, or reciprocal.

Ehrlich and Raven envisioned a plant lineage evolving a new defense trait to “escape” herbivory and “radiate” into a clade of daughter species carrying the new defense, until a herbivore species evolves a counter-defense and, in its turn, radiates into a clade of species feeding on the recently diversified plant resource. Adaptation occurs at all stages of “escape and radiate”, but it need not be reciprocal, much less simultaneous, or between specific pairs of plants and herbivores. Plants can “escape” whether or not they exert selection on attacking herbivores, because herbivory in general creates selection favoring a novel defense. Similarly, the later radiation of herbivores need not exert specific selection on their host plants — they may simply adapt to the diversity of host plants that becomes accessible when they acquire their new counter-defense.

So we have specific, simultaneous, reciprocal adaptation that isn’t called coevolution (by Pimentel); and phylogenetic patterns that are called coevolution (by Ehrlich and Raven) but don’t require specificity, simultaneity, or reciprocity. Contrasting those papers and their motivations is clarifying. Pimentel demonstrates that it’s possible to think perfectly coherently about specific, simultaneous, reciprocal adaptation without giving it a special name. Ehrlich and Raven, on the other hand, demonstrate that species interactions can shape biodiversity in ways well beyond specific, simultaneous, reciprocal adaptation.

The modern commitment to a single sense of coevolution emerges from Daniel Janzen’s bracingly brief 1980 letter to the journal Evolution, in which he laid down what multiple generations of coevolutionary biologists now take as law: “‘Coevolution’ may be usefully defined as an evolutionary change in a trait of the individuals in one population in response to a trait of the individuals of a second population, followed by an evolutionary response by the second population to the change in the first.”

Janzen (1980) is the citation given for most iterations of the strict-sense definition you will read, from peer-reviewed articles to textbooks. It is mainly concerned, however, with what does not count as coevolution: species whose traits are complementary (because the traits may have determined which species can interact, rather than the interaction selecting for the traits), and species that exert selection on parasites or symbionts without having been shown to experience reciprocal selection from those associates. Janzen sets his definition up against Ehrlich and Raven, whose 1964 paper is the letter’s only citation, and who both appear in its Acknowledgements.

However, Janzen (1980) doesn’t actually specify that coevolution is adaptation, much less simultaneous adaptation! This is, I think, a symptom of an era in which biologists still often conflated evolution with natural selection. Beyond the scope of the 1980 letter, it seems clear Janzen was indeed thinking of adaptation, if not necessarily simultaneity. His writings on ant-acacia mutualism — his first big foray into coevolution in 1966 and popular writing in 1984, as useful bookends to the 1980 letter — discuss acquisition of adaptive traits in a phylogenetic, rather than population-level, perspective. That is, they focus on traits that are derived and useful for the interaction, not whether those traits are presently under selection.

Acacias grow swollen, hollow thorns to shelter bodyguard ants, and have extrafloral nectaries and “Beltian bodies” to feed them; the ants’ patrolling behavior and colony structure make them better suited to guarding acacias. These traits could be acquired in a series of innovations as a single lineage of ants and acacias became better and better adapted to mutualism — but they could also be acquired as ants and acacias swap their associations between better and better partner species. That sounds almost like escape and radiate, but for mutualism! Janzen walks right up to the edge of that scenario in the Discussion of the 1966 paper, hypothesizing that once one ant/acacia pair starts down the road to mutualism, co-occurring acacias would benefit from adapting to host ants themselves.

Bodyguard ants (genus Crematogaster) crawl over the twigs, thorns, and a domatium formed by a swollen thorn of a whistling thorn plant (Vachellia drepanolobium). (Forestry Images, Whitney Crenshaw)

I find coevolution defined as specific, simultaneous, reciprocal adaptation not in Janzen’s publications, but in the 1983 book Coevolution, in the first paragraph of the Introduction by its editors, Douglas Futuyma and Montgomery Slatkin. Futuyma and Slatkin cite Janzen (1980) as providing a “restrictive definition”, relative to Ehrlich and Raven, and read Janzen as requiring “specificity” and “reciprocity” — one interacting species’ trait must exert selection on the trait of the other, and both traits must evolve. They then propose, tentatively, to out-Janzen Janzen: “A still more restrictive definition would also require simultaneity — both traits must evolve at the same time.”

In practice, the definitions of coevolution employed in the book vary with the contributing authors, and Futuyma and Slatkin structure the Introduction nicely by discussing how different definitions align with the real dynamics of species interactions — are a Batesian mimic and its model “coevolving” if their matched patterns are ultimately under selection by predators? — and the kinds of evidence we might use to address whether two species are coevolving. They circle in, however, on a preference for the restrictive definition of reciprocal adaptation, because “Coevolution, too broadly defined, becomes equivalent to evolution.”

Why do we care whether coevolution is subsumed into the rest of the field? Futuyma and Slatkin say coevolution deserves distinction “primarily because it is a major point of contact between evolution and ecology.” Biotic interactions are daily facts of ecological life that are, at the same time, linked to key innovations and major radiations across the Tree of Life. Although there are examples of radiations and key innovations evolved in response to the non-living environment, “many evolutionary events must be understandable only in the context of [trophic, competitive, antagonistic, and mutualistic] interactions.”

That latter point is, I think, indisputably true. Or at least, if I didn’t believe species interactions were important in the history of life, I’d have chosen a different subfield. But at this moment of agreement with Futuyma and Slatkin (and Janzen) I find myself back at my original question: do we need strict-sense coevolution to understand how species interactions have shaped the history of life? Futuyma and Slatkin themselves describe all sorts of evolutionarily important outcomes from species interactions that need not create reciprocal adaptation, much less specific, simultaneous, reciprocal adaptation.

This pattern — of laying out strict boundaries between what is and isn’t coevolution, then ranging right past those boundaries — carries on into discussions of coevolution decades later. John Thompson taxonomized “concepts of coevolution” in a 1989 review to which I’m indebted for the Pimentel (1961) reference, and then in books published in 1994 and 2005, and indeed in many papers, reviews, and syntheses over a field-defining career. Thompson’s geographic mosaic theory of coevolution puts specific, simultaneous, reciprocal adaptation (in the 1989 review, he dubs this “specific coevolution”) front and center, but embraces a broader range of processes as well. Over a geographic mosaic of populations on a variegated landscape, some host populations might escape a parasite species that afflicts them in other locations; some might face a different parasite species that interacts with the same resistance traits as the first; some might compete with a second host species that, in places where it is also hampered by parasite infections, is a weaker competitor.

There are near-infinite possibilities for coevolution to play out on a geographic mosaic, of course, and they are that much more difficult to demonstrate empirically — as well as that much easier to speculate about in the absence of complete data. The geographic mosaic perspective holds that any study of strict-sense coevolution must account for population genetic structure, community ecology, geographic variation in climate and nutrient availability — almost any other ecological or evolutionary factor that might impact the focal species interaction. Futuyma and Slatkin’s 1983 concern about “coevolution, too broadly defined,” might rear its head here, but really the geographic mosaic is about how coevolution interacts with other evolutionary processes. Its focus on interactions playing out across a geographically distributed metapopulation also places it neatly at the point where evolution within populations contributes to patterns that might eventually show up in the geological record — speciation, arising as different populations follow different local trajectories far enough to achieve reproductive isolation.

This is why I would have said strict-sense coevolution is necessary and interesting, if you’d asked me at any time from the beginning of my academic career — when my dissertation advisor handed me a copy of Thompson’s The Geographic Mosaic of Coevolution as introductory reading — to the tenure track. Speciation is an outcome of evolution across geographically distributed populations. Specific, simultaneous, reciprocal adaptation between interacting species seems particularly likely to set geographically separated populations on the path to speciation. If we want to know how species interactions have shaped biodiversity, we need to identify cases of specific, simultaneous, reciprocal adaptation between species. Q.E.D.

Yet this is not what we’ve demonstrated. The number of cases in which we can feel confident we’ve seen specific, simultaneous, reciprocal adaptation has grown, but the cases we can link to local adaptive divergence, much less speciation, really hasn’t. Reviewing the literature in 2014, David Hembry, Kari Goodman, and yours truly noted a “lack of widespread evidence for coevolutionary diversification” — that is, the evolution of reproductive isolation driven by strict-sense coevolution. The same year, David Althoff, Kari Segraves, and Marc Johnson surveyed evidence not just for coevolution-driven speciation, but coevolution-driven elevated rates of species formation — and also found “that evidence supporting an association between reciprocal natural selection and increased diversification is weak.” More recently, Anurag A. Agrawal and Xuening Zhang described “the link between reciprocal adaptation and diversification” as a “persistent gap” (ugh) in our knowledge.

This leaves us suspended between, on the one hand, evidence of many, many ways that species interactions can shape evolution, and on the other hand, almost no basis to say that coevolution is responsible for patterns of diversification and extinction at the scales of phylogenetics and the fossil record. A brand new review of the mutualism between yuccas and their hyper-specialized pollinators, yucca moths, by Chris I. Smith and Jim Leebens-Mack, sums up “15o years of coevolution research” only to arrive at the admission that no one has measured reciprocal selection in a yucca moth and its hosts, so we cannot really, definitely say that a clade of plants is coevolving with their exclusive pollinators. (Full disclosure: Chris Smith’s request for my comments on a draft of the “coevolution” section of that review sparked a reading binge that eventually became this essay, though I hope he won’t read it as directed at him personally so much as the subfield in which we both work.)

Western Joshua tree, Yucca brevifolia, in bloom this past spring near Mojave, California. (Flickr: jby)

Except for rare cases in which they’ve been transplanted outside their native range, yuccas have no other effective pollinators than yucca moths. Yucca moths, which lay eggs in the flowers they pollinate, have no other larval food source than those flowers. As Smith and Leebens-Mack review, a number of derived traits in both yuccas and yucca moths support this codependency — and as I read Janzen (1980) and Janzen’s other writing on coevolution, that sure should meet the standard to say yuccas and yucca moths have coevolved. Yet under the predominant definition of coevolution, it doesn’t.

This seems to me not so much a problem with our understanding of the yucca-moth interaction as a problem with our definition.

Here is my proposal: rather than focusing on whether a particular case study meets the criteria of a restrictive definition, let those of us who study coevolution consider all of the ways that interacting species can co-evolve in the most literal sense of the term, evolving together. Evolving, as I hope we can all agree, means more than adaptation! And together, as Ehrlich and Raven show, means more than specificity, simultaneity, or even strict reciprocity. (I’d argue this is also shown by Janzen, and Thompson, and many others, including Judie Bronstein, May Berenbaum, Risa Sargent, and Marc Johnson.) I’m calling, in short, to replace “is it coevolution?” with two questions: First, have species interactions contributed to the evolution of Earth’s biodiversity? And second, how did they do it?

I think we already have strong sense that the answer to the first question is “yes”. Phylogenetic studies, from Ehrlich and Raven’s to classic sister-clade analyses to modern phylogenomics, all address this, as do syntheses that compare biotic and abiotic sources of natural selection at the population scale, and population genomic tests for recent adaptive evolution. The second question is where things get interesting, because there are so many ways species might influence each other’s evolutionary history without (necessarily) creating specific, simultaneous, reciprocal adaptation. Here’s a few that occurred to me as I assembled this essay:

Species interactions shape the landscape of ecological opportunity. New mutualists offer new resources or services; escape from antagonists or competitors often means access to new resources and habitats. In other words, species interactions are closely tied to ecological opportunity, the opening of new “adaptive zones” that can support new biodiversity. This is true whether or not the mutualists, antagonists, or competitors experience specific, simultaneous, reciprocal adaptation. If interacting clades have created — or limited — ecological opportunity for each other, have they not evolved together?

Species interactions create new possibilities for reproductive isolation. Animal pollinators’ behavior shapes reproductive isolation between the plants they visit whether or not the plants and pollinators experience specific, simultaneous, reciprocal adaptation. Specialized symbionts — especially obligate, vertically transmitted ones — will often come to have phylogenies that closely match those of their hosts whether or not they exert meaningful selection on those hosts, just because the hosts’ phylogeny is effectively a series of vicariance events for the symbionts. Reproductive isolation is as much a process of evolution as natural selection, so if one species’ adaptation to a second species creates opportunities for isolation in that second species, are the two not evolving together?

Specific, simultaneous, reciprocal adaptation of interacting species can directly promote speciation. As discussed already, demonstrating this is a substantial task, and we have relatively few well documented examples; but it is clearly possible, and it’s exciting when we find it in operation. It’s likely that different types of interaction, which generate different forms of selection, may have different propensities to promote ecological divergence and the evolution of reproductive isolation. This is the case that already matches the consensus definition of coevolution, and the species to which it applies are unmistakably evolving together.

Even at its most inclusive, this approach still avoids expanding the definition of coevolution into a synonym for evolution. We can’t say that two species are evolving together if we haven’t identified how each species shapes the other’s evolution — that means showing they currently impact each other’s population dynamics, or dispersal, or survival and reproduction, or that they’ve done so in their shared history together. Janzen’s original approach of documenting derived traits that play a role in the interaction works here; so does finding significant parallels in phylogenetic structure, or linking the origin of partnerships to increased net diversification or changing trait evolution on a phylogeny, or finding signatures of adaptation in the genomic diversity of interacting species — or, again, estimating significant reciprocal natural selection in interacting populations. All the different kinds of evidence we use as evolutionary biologists can come into play, not just measures of selection gradients.

The secret of the coevolution literature is that everyone is already doing this. In paper after paper, we cite Janzen (1980) and the need to prove specific, simultaneous, reciprocal adaptation — then as often as not we fail to prove this and move right on to documenting one or more other ways that our favorite species interaction influences the evolution of the participants. To (maybe) make up for my cranky focus on Smith and Leebens-Mack (2024), I’ll say that I actually appreciate that their discussion of yucca-moth coevolution itemizes all the possible ways yuccas and moths could meaningfully evolve together. They find that in many of senses of coevolution, yuccas and their pollinators have most definitely coevolved. In this way, Smith and Leebens-Mack (2024) recapitulate Futuyma and Slatkin (1983), just within the scope of a single special mutualism.

Using coevolution in an inclusive sense would let us move past a narrow definition’s checklist of criteria — some of which aren’t even present in the widely cited source for that definition! — and get down to studying the myriad ways species can evolve together. Isn’t that, in the end, what we all came to this field to do?


Agrawal AA and X Zhang. 2021. The evolution of coevolution in the study of species interactions. Evolution. 75(7): 1594-1606.

Althoff DM, KA Segraves, & MT Johnson. 2014. Testing for coevolutionary diversification: linking pattern with process. Trends in Ecology & Evolution. 29(2), 82-89.

Berenbaum MR, C Favret, and MA Schuler. 1996. On defining “key innovations” in an adaptive radiation: Cytochrome P450s and Papilionidae. The American Naturalist. 148: S139-S155.

Briscoe Runquist RD, AJ Gorton, JB Yoder, NJ Deacon, JJ Grossman, Shan Kothari, MP Lyons, SN Seth, P Tiffin, and DA Moeller. 2020. Context dependence of local adaptation to abiotic and biotic environments: A quantitative and qualitative synthesis. The American Naturalist. 195(3): 412-431.

Bronstein JL. 2001. The exploitation of mutualisms. Ecology Letters. 4: 277-287.

Ehrlich PR and PH Raven. 1964. Butterflies and plants: A study in coevolution. Evolution 18(4): 586-608

Epstein B, LT Burghardt, KD Heath, MA Grillo, A Kostaneck, T Hämälä, ND Young, and P Tiffin. 2023. Combining GWAS and population genomic analyses to characterize coevolution in a legume-rhizobia symbiosis. Molecular Ecology. 32: 3798–3811.

Farrell BD. 1998. “Inordinate Fondness” explained: why are there so many beetles? Science. 281(5376): 555 – 559.

Flor HH. 1956. The complementary genic systems in flax and flax rust. Advances in Genetics. 8: 29-54.

Futuyma DJ and M Slatkin. 1983. Coevolution. Sinauer Associates, 555 pages. Google Books

Gomulkiewicz R, DM Drown, MF Dybdahl, W Godsoe, SL Nuismer, KM Pepin, BJ Ridenhour, CI Smith, and JB Yoder. 2007. Do’s and Don’ts of testing the geographic mosaic theory of coevolution. Heredity 98(5): 249-58.

Grant V. 1949. Pollination systems as isolating mechanisms in angiosperms. Evolution. 3(1): 82-97.

Hargreaves AL, RM Germain, M Bontrager, J Persi, and AL Angert. 2020. Local adaptation to biotic interactions: A meta-analysis across latitudes. The American Naturalist. 195(3): 395-411.

Hembry DH, JB Yoder, and KR Goodman. 2014. Coevolution and the diversification of life. The American Naturalist. 184(4): 425-38.

Hembry DH and MG Weber. 2020. Ecological interactions and macroevolution: A new field with old roots. Annual Review of Ecology, Evolution, and Systematics. 51: 215-243.

Janzen DH. 1966. Coevolution of mutualism between ants and acacias in Central America. Evolution. 20(3): 249-275.

Janzen DH. 1980. When is it coevolution? Evolution. 34(3): 611–612.

Janzen DH. 1984. The most coevolutionary animal of them all. Crafoord Lectures. The Royal Swedish Academy of Sciences. Available at

Johnson MTJ and JR Stinchcombe. 2007. An emerging synthesis between community ecology and evolutionary biology. Trends in Ecology and Evolution. 22(5): 250-257.

Kiester AR, R Lande, and DW Schemske. 1984. Models of coevolution and speciation in plants and their pollinators. The American Naturalist. 124(2): 220-243.

McKenna DD et al. 2019. The evolution and genomic basis of beetle diversity. PNAS. 116(49): 24729 – 24737.

Mode CJ. 1958. A mathematical model for the co-evolution of obligate parasites and their hosts. Evolution. 12(2): 158-165.

Pimentel D. 1961. Animal population regulation by the genetic feed-back mechanism. The American Naturalist. 95(881): 65-127.

Sargent RD. Floral symmetry affects speciation rates in angiosperms. Proc. R. Soc. Lond. B. 271: 603 – 608.

Schaller CW, Briggs FN. Inheritance of resistance to mildew, Erysiphe graminis hordei, in the barley variety, black Russian. Genetics. 1955 Jul;40(4):421-8.

Smith CI and JH Leebens-Mack. 2024. 150 years of coevolution research: Evolution and ecology of yucca moths (Prodoxidae) and their hosts. Annual Review of Entomology. 69: 375-391.

Thompson JN. 1989. Concepts of coevolution. Trends in Ecology and Evolution. 4(6): 179-183.

Thompson JN. 1994. The Coevolutionary Process. University of Chicago Press, 376 pages. Google Books

Thompson JN. 2005. The Geographic Mosaic of Coevolution. University of Chicago Press, 400 pages. Google Books

Thompson JN. 2013. Relentless Evolution. University of Chicago Press, 499 pages. Google Books

Yoder JB, E Clancey, S Des Roches, JM Eastman, L Gentry, WKW Godsoe, T Hagey, D Jochimsen, BP Oswald, J Robertson, BAJ Sarver, JJ Schenk, S Spear, and LJ Harmon. 2010. Ecological opportunity and the origin of adaptive radiations. Journal of Evolutionary Biology. 23(8): 1581-96.

Yoder JB and SL Nuismer. 2010. When does coevolution promote diversification? The American Naturalist. 176(6): 802-17.

Yoder JB. 2016. Understanding the coevolutionary dynamics of mutualism with population genomics. American Journal of Botany. 103(10):1742-52.

About Jeremy Yoder

Jeremy B. Yoder is an Associate Professor of Biology at California State University Northridge, studying the evolution and coevolution of interacting species, especially mutualists. He is a collaborator with the Joshua Tree Genome Project and the Queer in STEM study of LGBTQ experiences in scientific careers. He has written for the website of Scientific American, the LA Review of Books, the Chronicle of Higher Education, The Awl, and Slate.
This entry was posted in adaptation, Coevolution, ecology, evolution, mini-review and tagged , , , , , . Bookmark the permalink.