Can species arise in a small space?

August 6, 2011 • 10:51 am

One of the big controversies in the study of speciation involves the spatial scale of the process.  Can an ancestral species split into two descendants within a single small area (“sympatric speciation”), or do populations have to be geographically isolated before they can evolve into new species (“allopatric speciation”)?  Clearly the formation of new species is easier if gene flow between the incipient species is prevented, for that gene flow impedes the evolutionary divergence that creates new species. (My—and most biologists’—notion of species are groups that cannot exchange genes because of evolved, genetically-based barriers to gene flow.  Those “reproductive isolating barriers” evolve more easily in populations that are physically prevented from exchanging genes, such as those isolated by rivers or mountain ranges, or those that invade a remote island.)

In the book on speciation I wrote with Allen Orr in 2004, we concluded that sympatric speciation, while theoretically possible, was rare.  There were a few putative examples of species that seem to have arisen in small areas (the tiny lakes in volcanic craters of Cameroon, for example, contain more than a half-dozen species that descended from a common ancestor), but these examples were not common.

However, even if sympatric speciation were common, it would be hard to find, for you have to identify closely related species that you know were never geographically isolated. Given the history of climatic change and the movement of species’ ranges even in recent times (glaciation moves species around willy-nilly), that’s a tall order.

In our book, Allen and I suggested a way around this problem: look for closely related species confined to very small areas like isolated island and volcanic crater lakes.  If in those places you can find “monophyletic” species—that is, species more closely related to each other than to other species—you could reasonably conclude that those species had formed in those small areas—that is, that sympatric speciation had occurred.  Such studies don’t require one to know the biogeographical history of the species, since, living in small confined areas, they could never have been geographically isolated.

My colleague Trevor Price and I did the first systematic study of this problem, looking at bird species on isolated “oceanic islands” (that is, those islands that arose from beneath the sea, bereft of life).  Surveying 46 of these islands, we found not a single example of two avian “sister species” (i.e., each other’s closest relatives) on an oceanic island. Our conclusion: birds did not undergo sympatric speciation.  The same conclusion came from a survey of island lizards by Jon Losos and Dolph Schluter in 2000.  In these groups, then, geographic isolation seems necessary for the origin of new species.

In a new paper in The Proceedings of the National Academy of Sciences, Alex Papadopulos and six other authors continue this strategy, but looking at plants instead of vertebrates.  They chose to survey the flora of an isolated oceanic island, Lord Howe.  Lord Howe is small (about 16 square kilometers), was formed as a volcano about 6.9 million years ago, and is located between Australia and New Zealand:

Although I’ve never been there, it’s a gorgeous place. Here’s a aerial photo sent to me by one of the paper’s authors (click for a splendid enlargement). You can see that while a bit of of the island is settled, a lot of it is native forest, and thus affords a good chance to see if an ancestral species can split into two or more descendants in this small space.  An earlier paper by Savolinen et al. had shown that two species of palm trees on the island were each other’s closest relatives, and this result spurred scientists to do a more extensive survey.

So did they find speciation events occurring in Lord Howe plants?  The short answer is yes.  I’ve written a two-page commentary for the journal, explaining the results and why they are important. Although the paper itself and my commentary are behind a paywall, I’ll be glad to send pdf files to anyone who requests them by email.

A survey of the plants, combined with a phylogenetic study of the genera of those plants (such a study is required to show that two closely related species on Lord Howe really are each other’s closest relatives) identified at least nine other species—and perhaps as many as 18—that may have descended from common ancestors on the island.  Adding the two species of palms that were already identified, this brings the proportion of total species on the island that arose via sympatric speciation to between 4 and 8%. That figure is larger than most biologists would have guessed.

As I note in my commentary, this is not only an important finding, but one that can be extended to many other islands and groups.  After all, the oceans are full of isolated islands, and many of them have endemic species just begging to be studied systematically.  It’s through that kind of work that we’ll eventually learn how common is the process of sympatric speciation.  If it seems to occur often on these islands, then it probably also occurs often on the continents, where it’s much less likely to be identified.

Oh, and here are the two sister species of palms that formed on Lord Howe: Howea forsteriana (l.) and H. belmoreana (r.).  The former, also known as the Kentia Palm, is grown throughout the world as an ornamental plant.


Coyne, J. A. 2011.  Speciation in a small space. Proc. Nat. Acad. Sci. USA, published ahead of print August 3, 2011, doi:10.1073/pnas.1110061108
Coyne, J. A. and H. A. Orr. 2004.  Speciation.  Sinauer Associates, Sunderland, MA.
Coyne, J. A., and T. D. Price. 2000. Little evidence for sympatric speciation in island birds. Evolution 54:2166-2171.
Losos, J. B., and D. Schluter. 2000. Analysis of an evolutionary species-area relationship. Nature 408:847-850.
Papadopulos, A. S. T. et al.. 2011.  Speciation with gene flow on Lord Howe Island.  Proc. Nat. Acad. Sci. USA, published ahead of print July 5, 2011, doi:10.1073/pnas.1106085108
Savolainen, V., M. et al. 2006. Sympatric speciation in palms on an oceanic island. Nature 441:210-213.

49 thoughts on “Can species arise in a small space?

  1. Interesting, but unless the methodology can be transferred and used in other geographic regions, perhap this is an anomaly rather than a trend? I’d just caution to beware “the statistics of small numbers” and then extrapolate them into wider conclusions.

    1. What statistic would make you reject this as noise and accept it as signal?

      If it is a process, a binomial test would be too harsh I think.

      1. Wait, what are the chances that a phylogeny is not valid? Say 95 %. Then the likelihood that this is random grows astronomically: p < ~ (1-0.95)^11.

  2. Clearrly the formation of new species is easier if gene flow between the incipient species is prevented, for that gene flow impedes the evolutionary divergence that creates new species.

    So in a small space what could be the plausible mechanisms that prevent gene flows between incipient species? Could it be that some random mutation offers some selective advantage to some members but as a side effect also prevents or reduces the chance of breeding with other non-mutated members of the parent species?

  3. Jerry

    A most interesting piece, as usual, thanks.

    However, I just gotta ask for some clarification: is a visit from you usually required to render a place gorgeous or are you simply known for attempting to visit every gorgeous place on the face of the earth?


    1. I think you are reading that wrongly, Coyne is taken from a photo/trustworthy description.

    2. That reading, while grammatical, is most unnatural. Language is there to convey meaning. If anyone actually thought Jerry were making the island’s beauty contingent on his presence, rather than using a few short words to indicate that he’s not reporting said beauty from his own personal experience, you might have a point.

      1. Yes, of cours, that’s what I meant. How could anybody think otherwise unless they were a captious pedant? It’s clear that what I meant is that although I’ve never been there (but would like to), I know it’s gorgeous because I’ve seen photos.

  4. For plants, it’s not so difficult to think of a scenario for sympatric speciation.

    Consider a plant species S that is initially pollinated by two different kinds of animals, say a bee and a bird. If there is enough variability in the morphology and colour of the flowers of different individuals, then one can imagine that some flowers are more attractive to bees, and others more attractive to birds. Over time, birds would select for the bird-friendly forms, while bees would select for the bee-friendly forms. As a result, the bird-friendly forms would become even more bird-friendly, while the bee-friendly ones would become even more bee-friendly. At some point, birds would not be interested in the bee-friendly forms, while bees would not be intested in the bird-friendly forms. If you also assume that the original, intermediate forms would be less attractive to both birda and bees than the specialized forms, then you will effectively get genetic isolation, and thus sympatric speciation. An analogous process would arise if different pollinators occur in different seasons, while the flowering period of S spans both seasons.

    1. That scenario is not as easy as you think, because it has to lead to complete disjunction of the pollinators through disruptive selection. And, in the Lord Howe case, many of the plants, like the palms, are in fact wind pollinated, so the scenario doesn’t apply. And it makes it even harder to envision sympatric speciation in something that’s pollinated by wind.

      1. I agree that wind pollination makes the situation more difficult to explain. In that case I have to fall back on the usual reply, which is that we need more data.

      2. How about if there were two different winds, a daytime sea breeze and an evening or night-time land breeze? Different plants might point in different directions to^h^h and thereby differentially take advantage of them. They’d be at different temperatures, leading to further divergence. If, as some do, they opened and closed at different times of day, one can imagine very complete separation.

        (I have no botany and no idea whether this is at all plausible.)

      3. Are there physical barriers on the island that would allow a wind-pollinated plant to become isolated for a time?

        Is it reasonable to suppose that when it comes to plants, what constitutes geographic isolation is a good bit less strict than for animals, which can move about before reproducing?

  5. In my area of Ecuador, I see multiple examples of apparent sympatric speciation in the orchids I study. In my 60 km x 60 km study area, I discovered 30 new species of orchids of the genus Teagueia, and all of them are more closely related to each other than to the six previously known Teagueia species (we know this from both morphology and DNA analyses of multiple gene regions). There are 16 of these Teagueia on one single mountain. I also discovered weaker local radiations in the genus Lepanthes, including one set of five or six currently-sympatric new endemic species which are each other’s closest relatives as evidenced by morphology.

    I also looked closely at the mathematical basis for the common belief among evolutionary biologists and geneticists that the exchange of a few migrants per generation keeps populations from diverging. This idea (the “one-migrant-per-generation” rule”) turns out to be wrong. It is based on a mistaken measure of divergence, Fst. The thing that actually controls divergence at a locus (in simple neutral genetic models) is the ratio of relative migration rate to mutation rate. So two large populations could exchange dozens of their members each generation and still diverge strongly, even in the absence of natural selection. It is not true that isolation is needed for divergence. The work of Gavrilets showed that this same ratio controls the accumulation of genetic incompatibilities that can lead to reproductive isolation.

    1. Do you know anything about their pollinators?

      I know that Lepanthes species are pollinated through pseudocopulation by small diptera. Part of the flower even resembles insect genitalia. Given this very specic relationship, and the immense number of species of diptera, there appears to be ample scope for speciation based on pollinator specialisation. See also my previous comment.

      1. Yes, I suspect we can get nearly instant reproductive isolation in many orchids (especially those pollinated by pseudocopulation) just by making small changes to the flower or its fragrance. Lepanthes actually generate fake insect pheromones, apparently species-specific. No one knows what pollinates Teagueia. To my eye they are fairly unspecialized flowers, compared to Lepanthes.

        But whether these are cases of sympatric speciation or not, my main point was that complete isolation is not needed in order to get large amounts of genetic divergence (even in the absence of differential natural selection). Divergence at a locus is the norm when pairwise migration rate divided by mutation rate is low, evcn if there are many individual migrants per generation.

        1. Most mutations are either neutral or deleterious. Within the population there will either be no selection for them, or there will be selection against them. Only when fitness-increasing mutations can arise and spread in a (sub)population without spreading to other populations, only then could you get speciation. Therefore you only need to consider these particular mutations. However, I don’t think it is meaningful to speak of a ‘rate’ of fitness-increasing mutations, because this ‘rate’ is likely to be modified by the mutations themselves. So it would be wrong to model it as something static which can be compared to a rate of migration.

          1. No, the process has a large stochastic component, and this stochastic component can only be understood by figuring out what happens in the absence of natural selection. That is why it is important to work out what happens at neutral loci. Even if there is no benefit to a new mutation, it can spread through a population, or through all populations of the species, through drift. The ratio of pairwise migration rate to mutation rate determines whether, on the average, the alleles end up being shared among all populations, or are restricted to just one. This can be proven mathematically and is confirmed by simulations. And what happens at neutral loci provides the baseline for what happens to loci that are subject to natural selection.

          2. Sorry, but I have to disagree. You can’t extrapolate just like that from neutral mutations to ones that are subject to selection. In particular, as I tried to explain, I don’t think it even makes sense to talk about a ‘mutation rate’ in the case of mutations that increase fitness. There is no simple, fixed stochastic process for these as there is for neutral mutations. A single fitness-increasing mutation M could render a large set of other potential, previously beneficial mutations harmful instead. At the same time, potential mutations that would have been harmful at first could become beneficial after M occurred.

            The conceptual difference between neutral mutations and beneficial ones is like the difference between the Special Theory of Relativity, where there is a fixed metric, and the General Theory of Relativity, where the metric is variable. The latter situation is vastly more complicated.

            Genetic drift mainly occurs in small fringe populations; it is very unlikely to spread across large populations throughout the whole distribution area of a widespread species. Only selection could cause that to happen. But then we are not talking about neutral mutations.

          3. Drosera, to continue your analogy with relativity: Just as general relativity must give special relativity as gravity becomes weak, the predictions of general theories of evolution must agree with those of the neutral theory as selection becomes weak.

            You say that drift occurs only in small populations and that new neutral mutations are unlikely to spread across all populations of a species. I respect your intuition, but this is not true. As I mentioned, the equilibrium divergence under the finite island model depends entirely on the ratio of pairwise migration rate to mutation rate. If the ratio is large, all populations will end up sharing all their common alleles. On the other hand, genetic divergence will arise without the aid of natural selection, even in large populations, if that ratio is small. These conclusions have been proven mathematically and confirmed by simulations. See the simulations in the Comment section following this post:


            I do not really understand your comment about mutation rate. The rate of new mutations arising at a particular locus is well-defined. Mutations don’t know they are going to be advantageous or deleterious in advance; this is a contextual quality determined partly by the environment. In this context, geneticists do not speak of a rate of fitness-increasing mutations, just the rate of mutations. We then follow them on their stochastic ride through the populations, assuming no selection. If differentiation arises stochastically even without divergent selection, or if populations become homogeneous through stochastic processes, that tells us something important about the general case when selection is active.

          4. Hi Lou. I’m afraid I don’t quite understand what you are saying either. It is of course trivial that in a certain population neutral mutations can arise that will become established within that population, but which will not spread easily to other populations when there is little gene flow between the populations. So over time, there will be genetic divergence between populations with respect to neutral sites. That is not speciation. You claim that in your model the probability that a neutral mutation becomes fixed in another population depends on the ratio between the mutation rate at that site and the migration rate. I haven’t seen your model, but I have no reason to doubt it.

            But as soon as the mutation is fitness-increasing, non-neutral, the situation becomes rather different, I would say. Then the likelihood that the mutation spreads to another population would merely depend on the minimum number of migrants needed to establish a foothold in this other population. Natural selection would favour carriers of the mutation, so it will quickly spread throughout the population. In this case, the absolute number of migrants would be the determining factor, not the ratio between migration rate and mutation rate.

            Once a site is no longer neutral, the way it changes over time is not purely random any more, because it is subject to selection. Its fate becomes dependent on that of other sites and even on the environment. Would such a site be a useful component of a molecular clock? I would say not. But then it is meaningless to talk about the mutation rate of that site. So, for non-neutral mutations it makes, in my view, no sense to talk about the ratio between migration rate and mutation rate. I hope I made myself clear this time.

            Otherwise we just keep trying until our columns are one letter wide 🙂

          5. Your last comments help me understand a bit. You have a few misconceptions about what I was saying, probably my fault.

            My model is the standard finite island model of population genetics. My observations also apply to the more realistic stepping stone model.

            It is not trivial to find the conditions that cause a new mutation at a neutral locus to spread to other populations. The ratio I mentioned provides the conditions. If you would look at the link I included above, you would see that the classical answer (the absolute number of migrants) is wrong. That this mistake went undetected for seventy years suggests the problem is not trivial.

            Also, I was never talking about probability of fixation. Two populations that each have lots of alleles (hence no fixation) can still be completely diverged, if none of these alleles are shared across populations.

            Third, of course this is not speciation, but speciation presupposes divergence at one or more loci. A neutral model is the null model for the speciation process. If selection is weak, this model will be close to the truth. Nothing about the process changes discontinuously as selection goes from 0 to some nonzero number.

            I do agree that non-neutral loci are poor choices for molecular clocks.

          6. Drosera, take your model (or any model) of what happens when a favorable allele arises. That model has to fade into a neutral model if the selection effects are very small. Just as general relativity gives special relativity when gravity is weak. The nice thing about this is that we know how to get analytical results for neutral models.

          7. Thanks for the clarification and the interesting discussion. I will look into your papers, as the link itself was not sufficiently detailed about your model.

            I agree that a neutral model is a limiting case under very low selection and as such useful for analysis. However, I think that speciation requires selection and/or isolation (except perhaps in cases of polyploidy, but I don’t think these are very frequent in orchids), so it is not straightforward to generalize from your model. But I see that you already realize that.

          8. Yes, I agree that generalization to cases of strong selection is not straightforward, but I disagree that speciation requires selection or isolation. Drift (even in the presence of gene flow) can create genetic differences as well, and this can eventually lead to reproductive isolation. I mentioned elsewhere on this website the work of Gavrilets on the evolution of speciation; he also emphasizes this point.

            Regarding my model, again, it is just the standard finite island model that everyone uses.

            One thing we have not discussed is how long it takes for drift to cause genetic divergence. Gavrilets shows that this is where selection makes a big difference.

  6. Here is an example of a Cyprinodon species flock in a Mexican lake.

    Sympatric speciation by means of polyploidy should be fairly common in plants. Also production of infertile hybrids which then undergo polyploidy to become fertile while still isolated from both parent species.

    I think there a situation in tree frogs in Florida where there are syntopic 2N and 4N species.

    In allopatric speciation there can be no selection for isolating mechanisms as such because the two incipient species are not in syntopic contact. In sympatric speciation (or after syntopic contact of incipient allpatric species) there would be selection for isolating mechanisms if individuals of the two incipient species who hybridize have reduced fitness.

  7. Jerry writes: My colleague Trevor Price and I did the first systematic study of this problem, looking at bird species on isolated “oceanic islands” (that is, those islands that arose from beneath the sea, bereft of life). Surveying 46 of these islands, we found not a single example of two avian “sister species” (i.e., each other’s closest relatives) on an oceanic island. Our conclusion: birds did not undergo sympatric speciation.
    I don’t understand this at all–what am I missing? I’m thinking of the Hawaiian honeycreepers, 50+ new species, half dozen or more new genera, an entirely new family (or sub-family, at least), evolved on an oceanic island group. Are there not many pairs of “sister species” within the Drepanididae/Drepanidinae?

    1. You’re missing that we looked at SINGLE and ISOLATED OCEANIC ISLANDS, not archipelagos. Hawaii is an archipelago, affording ample opportunity for allopatric speciation via inter-island migration.

      That’s what you’re missing.

  8. Well, it is all relative, isn’t it? Even on a small island like that, if the two emerging species just spent a few tens of thousands of years on opposite ends of the island, is that allopatric or still sympatric? With homoploid speciation not driven by a pollinator shift, I would assume that some manner of allopatry at some spatial level, even if a very fine one, is unavoidable.

    On the other hand, for plants at least it is trivial to think of two mechanisms that would allow sympatric speciation, alluded to above: Polyploidy gives instant genetic isolation (unless the parental diploid population becomes polyploid so often that it keeps pushing genes into the new polyploid one, but it would at least mostly be a one way street). And if one or maybe two very simply mutations can markedly change a flower in colour and shape, as has been demonstrated for some genera, then they alone may cause sufficient genetic isolation through a pollinator shift.

  9. Thinking of wind-dispersed plants, as a number have stated, one requires additional information; most important IMO would be measurements of variation in seed size, important because smaller seeds will travel further than larger ones (measurable, also). Thus, migration rate should be relatively easy to estimate (given a lot of sweat in the field). Assuming that these metrics are a small piece of the puzzle, in order to create a working model one would need to measure variations in microhabitat, whether the viability of seed sizes (say, small, med, lg) differs predictably x microhabitat type (easily doable), and (this is key, of course) whether wind direction is relatively constant during seed production season (easily measurable). A major “rub” (measurable) would be to determine how the med-sz seeds are behaving. Something similar to habitat selection may be going on w/sm & lg seeds leading to divergence. I agree w/one commenter that spatial scale (of island relative to seed sz) and features of microhabitat (relative to seed sz & viabilities) require measurement (both doable). If required to know mutation rate (x seed sz), I assume but don’t really know that these data would have to be collected in lab.

  10. Dr. Coyne,

    I’ve read in papers that social parasitism in ants arose from sympatric speciation because the social parasites themselves generally, are closely related to their host species(such as the genera Lasius and Formica).

    What are your thoughts?

    1. Allen Orr and I deal with this question in our book Speciation (p 154-155). The sympatric scenario for social parasitism, or “inquilism” is, as E. O. Wilson points out, very difficuult to envision, and he suggests a plausible allopatric alternative. Inour book we describe this as follows: “a free-living ancestor could speciate allopatrically and, after secondary contact, one species could become parsitic on the other. It is likely that parasitic ant queens could invade host colonies more easily if they are closely related species, so we might expect that inquilines and their hosts might often be sister species.”

      We don’t know for sure, but we can’t automatically say that inquilines must have arisen sympatrically.

  11. Thank you! Again a very interesting biology-post. These posts are my favourites! (Well – with boots and food as sharp competitors 😉 )

      1. Yes, I know what you meant. If you had smilies I would have included one with a friendly wink.

        I’m just saying that the way you said it sounds like when you go someplace it makes that place more beautiful.

  12. Here is a fascinating video of speciation in a small space – The Congo river has such a high flow rate that small sections of river bed are effectively isolated from each other, and new species have evolved in each successive microenvironment:

  13. It is an oversimplification to say that species cannot interbreed. Species ordinarily do not interbreed under usual circumstances. Habitat disruption, introductions, range expansions, mosaic ecotones, captivity and various other not usual situations can lead to formation of hybrid swarms which may persist for long periods. There is a large literature on hybrids among species.

    There are a number of lakes with species flocks. Someone said we would not fully understand speciation until we could explain species flocks in lakes.

  14. There are a two possibilities to rule out before making this conclusion on speciation.

    Lord Howe Island is rapidly eroding, it is now only 5% of the size it was size 6.7 million years ago. There may have been geological barriers back then.

    Lord Howe island is part of a chain of islands that are formed and eroded above a volcanic “hot spot” moving beneath the earth’s crust. Elizabeth and Middleton reefs are the remnants of older eroded islands. Geologists may be able to say how often islands are simiultaneously above sea level which would allow species to both island hop or evolve in isolation. Island hopping would be a rare event, the islands in the chain are separated by up to 100km.

    1. ^ How interesting to learn where this idea (that essentially sank this putative case of sympatric speciation) was first proposed!

Leave a Comment

Your email address will not be published. Required fields are marked *