How (and how fast) do new species form?

January 4, 2017 • 10:00 am

Most evolutionists think that speciation, which we see as the origin of a new group whose members are unable to produce fertile hybrids with other such groups (but whose members are interfertile with each other) occurs in the following way. Populations of a single species become geographically isolated by the interposition of geographic barriers like mountains, deserts, water, continental drift, etc. These barriers can either arise de novo, like the Andes, thus isolating populations on either side of them, or result from a rare migration event that is a one-off, like the invasion of the Galápagos by ancestral iguanas, tortoises, or finches. By preventing the mixing of genes among populations of what was initially a single species, the populations can then diverge genetically, often by natural or sexual selection, but also by processes like genetic drift. Different environments, different selective pressures and different mutations will then assure that the geographically isolated populations will travel along different evolutionary pathways.

When sufficient genetic divergence has occurred that the populations can no longer produce fertile hybrids if they were once again to come into contact, we say that they are separate biological species. The genetic barriers preventing successful hybridization, called reproductive isolating barriers, are diverse: they can include a preference for different environments, mating at different times (“temporal isolation,” which occurs in corals), a dislike of mating with the other species (“sexual isolation”, very common in birds and flies) or the production of either inviable or sterile hybrids (e.g., the sterile mule) when members of different species do mate. What has happened is that the geographic isolation allows the evolution of genetic isolation up to the point of speciation. 

The scenario I just described is called allopatric speciation (from the Greek meaning “different places”). But there is evidence that species can form without such geographic isolation, especially in the case of polyploidy in plants, in which two species hybridize (or a single species doubles its genome), and genetic processes in the hybrid make it a different species from either parent. This can take only a handful of generations. And we are learning that, more often than we thought, species can still split while exchanging some of their genes.

Now all of this represents our best current take about the origin of species, which is summed up in the technical book I wrote with Allen Orr in 2004, Speciation. These views were not developed by Darwin, as he didn’t have a clear idea of the relationship between species and reproductive isolation. Although his most famous book is called On the Origin of Species, it says little about that issue, but rather deals more with the origin of adaptations within a species.

But how long does it take for this process to occur? That’s harder, as only in very rare cases can we actually observe new species arising, and only when it’s fast—as in polyploidy (about 3 generations to get a new plant species). Without our having been there, we have to use indirect methods.  Allen Orr and I, in a pair of papers in Evolution (references below), tried to do that with the fruit fly Drosophila. 

What we did was take named species of flies whose degree of reproductive isolation—only sexual isolation and hybrid inviability and sterility—had been measured in the lab, and estimated the “genetic distance” between those species using early electrophoretic data. In that way we could correlate the degree of reproductive isolation of different species with the degree of genetic divergence between them. Assuming the latter statistic accumulates roughly linearly with time (the so-called “molecular clock”), we could then get an idea of how fast reproductive isolation accumulates. Our admittedly rough estimate—for we neglected forms of reproductive isolation impossible to measure in the lab (i.e., ecological differences)—was that it took between 200,000 and 2.7 million years for one species of Drosophila to split into two. (The faster estimates are for species now found in the same area, for in such cases natural selection against hybridization, if hybrids are sterile or inviable, can speed up the evolution of reproductive isolation.)

In our book Speciation, Allen and I considered many other groups, and concluded that in general speciation takes between 0.5 and 5 million years, with some exceptions like polyploidy. But that’s a very rough guess based on many kinds of data, including fossils. That is, by the way, ample time to generate the millions of species living today, even taking pervasive extinction into account.

A new paper in PLOS Biology by Camille Roux et al. (reference and free link below) uses a related method to see how much genetic distance between groups is sufficient to make them different species, although in this case the species were designated simply by nomenclature rather than by strict consideration of reproductive isolation. The methods are complex, but I’ll oversimplify them to give the general result.

Roux et al. used data on 61 pairs of species or populations; these were diverse, including worms, fish, crustaceans, mammals, and insects. All of these pairs had extensive molecular data known for them. Ten were pairs taken from the literature, 22 were pairs of named, distinct species, and 29 were pairs of populations considered to be within the same species.

Using various complex models, the authors attempted to see how much gene flow was going on between these pairs of taxa, relating that to the genetic distance between them. (There are ways to do this without it being circular.) They used various models, including complete allopatric speciation, speciation occurring while gene flow was going on between the separating groups, and gene flow that occurred after geographically isolated groups once again came to live in the same place. What they came up with is a plot showing the relationship between the degree of genetic difference between pairs of species or populations (Da below) and the probability of ongoing gene flow between taxa (P in the diagram below).

(From journal): Fig 3. Probability of ongoing gene flow along a continuum of molecular divergence. Each dot is for one observed pair of populations/species. x-axis: net molecular divergence Da measured at synonymous positions (log10 scale) and averaged across sequenced loci. y-axis: relative posterior probability of ongoing gene flow (i.e., SC, IM, and PAN models) estimated by ABC. Red dots: pairs with a strong support for current isolation. Grey dots: pairs with no strong statistical support for any demographic model (robustness <0.95). Blue dots: pairs with strong statistical support for genome-homogeneous ongoing gene flow. Purple dots: pairs with strong statistical support for genome-heterogeneous ongoing gene flow. Filled symbols: pairs with a strong support for genome-heterogeneous Ne. Open symbols: genome-homogeneous Ne. The light grey rectangle spans the range of net synonymous divergence in which both currently isolated and currently connected pairs are found (see S1 Data).

What does this diagram show? Well, as you move along the horizontal (X) axis from left to right, the genetic distance at “neutral” sites increases, meaning that the species or population pairs get older. And you see from the Y axis (gene flow) that as the age between taxa increases, the amount of gene flow between them decreases, finally winding up at zero—at which point they are full biological species (in red). The Big Deal about the paper, however, is the fairly narrow transitional “gray zone”: the bar that marks the transition between fairly free gene flow (populations or incipient species) and full species. That gray bar extends from 0.5% to 2% sequence difference at “neutral” sites.

Surprisingly, the transition zone is about the same regardless of whether the groups are geographically isolated or not; I would have expected that geographic isolation would speed up speciation by impeding “annoying” divergence-preventing evolutionary difference. The authors were also surprised at this, but said this may be because their range data were not so great. Further, the authors recognized several new species, which looked pretty much the same but whose genetic distance put them on the right (i.e., not left) side of the speciation threshold.

The fact that diverse taxa adhere to the same “threshold” of speciation seems surprising, implying that the amount of neutral genetic difference associated with speciation is roughly the same for very different groups. From that observation the authors conclude this (my emphasis):

. . . our report of a strong and general relationship between molecular divergence and genetic isolation across a wide diversity of animals suggests that, at the genome level, speciation operates in a more or less similar fashion in distinct taxa, irrespective of biological and ecological particularities.

This conclusion is echoed by a PLOS blog post on the paper by grad student Jenns Hegg from the University of Idaho:

Does this paper tell us what is and isn’t a species? No, it doesn’t. But, it gives us an idea of how to understand the process of speciation across species. It also indicates that speciation happens (genetically at least) in pretty similar ways in all species regardless of the specifics of the population…which is good news for anyone interested in developing better ecological and evolutionary theories to explain how species come about through natural selection.

Well, both of these conclusions are dubious. Yes, it was surprising that the “gray zone” wasn’t so different among diverse taxa, but remember that these taxa are not all phylogenetically independent. Some groups are used more than once, so we can’t be sure that every data point represents an evolutionarily independent pair of taxa. Further, the sample size is limited (e.g., no Drosophila!) Also, there are these problems:

  • We know of some cases of speciation that are very rapid, including about 4000 years in some cichlids, 50 years in polyploid plants, and a few hundred years in sunflowers (see here for some fast cases of speciation that are outside the gray zone). Polyploidy, or rapid ecological speciation, obviate these conclusions. One cannot conclude that all speciation events adhere to the pattern above.
  • A constant genetic distance of 0.5%-2% does not necessarily mean that species take roughly the same time to evolve in nature. That’s because the molecular clock ticks at different rates in different taxa, so a divergence of 1% in flies could represent a very different time for a divergence of 1% in mammals or worms.
  • We already know that different taxa speciate in different ways, so saying that “speciation happens in similar ways in all species regardless of the specifics of the population” is just wrong. (Polyploidy instantly invalidates that statement!) And it’s wrong to say, as the authors do, that “speciation operates in a more or less similar fashion in distinct taxa.” What the data show is that the neutral genetic distance associated with the reduction of gene flow is similar in this small sample of species. The paper doesn’t—and cannot—say anything about the process or “mechanism” of speciation, which we already know differs in different groups. We can’t even say that the rate of speciation is similar in different groups until we know how neutral genetic distance translates into years among different groups.

I think, then, that too much is being made of this paper, though the results are still quite interesting. But we need a much better data set, a much better calibration of the molecular clock in different groups, and some clearer thinking about what the authors mean by “speciation operating in a similar fashion.” We already know that in different groups different reproductive isolating barriers are important (pollinator isolation in orchids, sexual isolation in birds, etc.—see Speciation). And if by “similar fashion” the authors mean “similar rates,”  well, that remains to be seen as well.

The authors discuss the use of this gray-zone metric as a conservation tool—a way to distinguish taxa to put them within the regulations for protection of endangered groups—but I won’t open that can of worms.


Coyne, J. A. and H.A. Orr.  1989.  Patterns of speciation in Drosophila.  Evolution 43: 362-381.

Coyne, J. A., and H. A. Orr. 1997. “Patterns of speciation in Drosophila” revisited. Evolution 51:295-303.

Roux C, Fraïsse C, Romiguier J, Anciaux Y, Galtier N, et al. (2016) Shedding light on the grey zone of speciation along a continuum of genomic divergence. PLOS Biology 14(12): e2000234. doi: 10.1371/journal.pbio.2000234

76 thoughts on “How (and how fast) do new species form?

  1. I got a lot out of this post. Thanks.

    The intro on speciation was a great review. And while Fig. 3 may not hold up, I find it an interesting framework within which to think about things, even if exceptions will clutter it up.

  2. My understanding of polyploid speciation is not correct, I think. Is it not the case that a polyploid offspring is typically unable to successfully reproduce with members of its parents species? So that in one generation there is a complete barrier to gene flow*, i.e. a speciation event?

    *Not including bacteria and other microorganisms in which lateral gene transfer is significant.

    1. It takes two or three generations for a new allopolyploid species to form. The first generation is a hybridization. Then other things have to happen to the hybrid in subsequent generations.

      An autopolyploid could theoretically form in a single generation, but usually takes at least two. And yes, in both cases the new polyploid is unable to produce fertile offspring with its ancestor (a few can be fertile but polyploids often seem to occupy new ecological niches that further prevent mating).

      1. Do they occur mainly where humans have introduced plants that are related so they are in close proximity, or more in nature, not that I suppose they are common? Thinking of the ‘London’ plane, Platanus x acerifolia…

        Also PLEASE – when a plant has ‘x’ in the middle of the name, indicating a hybrid, how do you say it? i.e. ‘ex’? or maybe ‘cross’?

      2. I wonder if in ferns new polyploid species can arise in one generation. Infertile hybrids can exist as sporophytes, but they cannot produce good spores, and so no gametophytes grow. I think chromosome doubling must happen at the time of fertilization, which occurs between sperm and egg produced by gametophytes that have grown from spores produced by fertile sporophytes.

  3. …speciation, which we see as the origin of a new group whose members are unable to produce hybrids with other such groups (but whose members are interfertile with each other)…

    I wish you wouldn’t take this as the definition of speciation. At the least, it’s a poor fit to birds, which notoriously retain the ability to hybridize for many millions of years after separation into distinct groupings, and we would like to call those groupings “species”. It may be that birds are unique, though I doubt it. Many plant groups appear to behave similarly; consider Quercus, for example. You may qualify the definition above by talking about “good species”, but I think bird species are perfectly fine.

    I note that the paper didn’t assay any birds, as well as Drosohila.

      1. I would use the biological species concept of Mayr, which is not usually so draconian as Jerry’s statement of it. Reproductive isolation need not be complete, nor does it need to incorporate any post-zygotic or post-mating mechanisms.

        Mayr said “species are groups of interbreeding or potentially interbreeding natural populations that are reproductively isolated from other such groups”, and in practice he interpreted that loosely. He famously produced a classification of ducks, for example, a taxon in which fertile hybrids between genera are not at all uncommon, especially in captivity.

        1. Umm. . . how complete does it have to be before you call it speciation? If it’s less than 100%, you’re making a purely subjective judgment. Again, we discuss this in our book, saying that things can be “more or less species like”. I presume you’ve both read the book and know about this, so you’re not being fair when you said my concept is “draconian”. Allen and I clearly distinguish between “full biological species” and things that are more or less “species like,” depending on the degree of reproductive isolation.

          You may consider my concept draconian (it’s not), but yours is totally subjective.

          Let’s end this discussion now. Again, I refer reader to chapter 1 of Speciation.

    1. Sorry, John, but I justify my definition in my book; reproductive isolation is the explanation why nature comes in clumps rather than a continuum, which is, as the founders of the modern synthesis noted THE problem of speciation.

      If birds form distinct groupings, what other reason is there besides reproductive isolation. Yes, there can be hybridization, but the hybrids don’t thrive (reproductive isolation). If they do, there wouldn’t be distinct groups; there would be a continuuum.

      Take this up with my next-door colleague, evolutionary ornithologist Trevor Price, who wrote a book on bird speciation that also uses the reproductive-isolation concept.

      I wrote a book on speciation, and you’re telling me that you wish I wouldn’t use that definition of species? Seriously? I suggest you write your own book based on your own definition rather than come over here and carp about mine, which I’ve justified in extenso in Chapter 1 of my book with Orr.

      1. I think you are fudging the definition here. “Reproductive isolation” and “unable to produce hybrids” are quite different criteria. Reproductive isolation can occur, for example, purely by (genetically based) behavioral mechanisms, and those mechanisms need not be perfectly reliable. Avian hybrids are quite commonly both viable and fertile; they just aren’t very attractive to members of either parent population, and so there is strong selection against them, and they have minimal reproductive success.

        In your book, you adopted the biological species concept, but what you propose here is an extreme version of that concept, one that Ernst Mayr, who coined the term, would not accept. I like the BSC myself; we’re just quibbling over its precise meaning.

        1. With due respect, John, you don’t know what you’re talking about. EVERY FORM OF REPRODUCTIVE ISOLATION impedes the formation of fertile hybrids. Either the hybrids don’t appear, or they appear and are weak, or they appear and are sterile.I’ve added the word “fertile” in the text to make that clearer. Behavioral isolation, as you well know, makes species unable to produce hybrids, and, as you must know from our book, fertile but unattractive hybrids instantiate a form of intrinsic postzygotic isolation: they are, in effect, behaviorally sterile. The concept I propose here is the same as the one in the book, and I urge readers to consult chapter 1 of speciation if they want to see a fuller discussion, including the PROBLEM OF THE DUCKS. I’m not fudging anything here, and that’s an accusation of malfeasance for which you should apologize.

          1. He may have been in his 90s, but Mayr was pretty clear that lineages had to not be able to interbreed to be considered a species (22:07 – 22:46)

    2. Seems as if it would be an interesting test to apply their methodology to birds, then! Is the data available to do that?

      1. If the hybrid is less fit than either form, we have a hybrid breakdown, which is considered a postzygotic reproductive mechanism. Moreover, even if they can form fertile hybrids, if they don’t do in nature, we have prezygotic reproductive isolation in place and, hence, separate species.

  4. Why are rates in years important? I would lean towards generations, if talking about evolutionary time (as you mentioned, the molecular clock ticks differently in different lineages). But it seems like the authors are trying to address whether the amount of neutral genetic divergence predicts whether barriers to gene flow have accumulated (inferred from lack of evidence for migration). In this context, a “narrow grey region” is interesting, if it’s interpretable. Obviously there are exceptions that would fall outside this grey region, as you note. Which taxa, other than polyploid plants and Drosophila, are missing from their 61 pairs? What kind of sampling scheme would allow better conclusions about the time-scale (or genetic distance scale) of speciation?

    1. Year rates are important just because people want to know. But we also don’t know if the molecular clock ticks at a given rate across taxa in generations. And as for your interpretation above, yes, that’s one of the things that authors are trying to do, but they are also trying to infer whether speciation “works” the same way as well as make other conclusions about conservations.

      If you want to know what taxa are missing from the 71, just download the dataset. You are capable of doing that, aren’t you? I don’t want to have to do it for you; suffice it to say that many groups aren’t represented.

      1. “we also don’t know if the molecular clock ticks at a given rate across taxa in generations” – thanks – answers my question!

      2. Yes, anyone who wants to download their species list can do so; it’s “S1 Data.” I find it interesting that they don’t include it as a main table. Why these species? (I know they cite Romiguier et al. to justify their choices, but I can’t access that paper). They’re obviously not the only pairs with genomic data available. If one of the criticisms of previous works is that they’re “limited case studies,” one should also be mindful of the limitations of the present work. I’m curious about how future studies can do a better job of sampling to allow comparisons across taxa with larger sample sizes.

        1. For following a little bit the speciation litterature, I am shocked by this criticism.
          It’s exceptionnal to see such amount of data, obtained in an homogeneous framework, analyzed in the same statistical framework, with species which are not historical models for a specific mode of speciation.
          A large body of the litterature wants to make a scoop by finding cases of “speciation in face of gene flow”, because it has been proven to be more difficult to occur than the simple evolution of barriers between isolated lineages, and so, expected to be more rare. Finding something rare brings prestige, but it doesn’t make frequent those rare cases…It’s like visiting Fort Knox and saying: “after all, gold is frequent on Earth”.
          To appreciate special/exceptional cases of speciation, it may be important to know before what is the general picture, no?

  5. Very interesting. Thanks for thaking the time to share this.

    Surely though, speciation could take a whole lot less time in small fast reproducing creatures like flies than in say humans or elephants?

    PS This on bird song & speciation may be of interest-
    Song evolution, speciation, and vocal learning in passerine birds.
    “Phenotypic divergence can promote reproductive isolation and speciation, suggesting a possible link between rates of phenotypic evolution and the tempo of speciation at multiple evolutionary scales. To date, most macroevolutionary studies of diversification have focused on morphological traits, whereas behavioral traits including vocal signals are rarely considered. Thus, although behavioral traits often mediate mate choice and gene flow, we have a limited understanding of how behavioral evolution contributes to diversification. Furthermore, the developmental mode by which behavioral traits are acquired may affect the rates of behavioral evolution, although this hypothesis is seldom tested in a phylogenetic framework.”

  6. Fascinating. In time these questions of speciation should resolve as more data and more studies pile up. Makes me want to live a lot longer, just out of curiosity.

  7. Neat. I will have to take time to digest.

    However, one comment: 50 years! That’s *just* within the possibility of being noticed within the career of one scholar – like, say, Richard Lewontin … (or Ernst Mayr, while he was alive). That’s nifty.

      1. According to Speciation by Coyne and Orr 2004, non-sexual reproduction can result in discreet species through “recurrent episodes of periodic selection.” New mutations can arise in non-sexual organisms and be passed down in separate lineages creating differing clusters. Periodically a new mutation is adaptive and the clone with this mutation “will replace all other clones with which it is ecologically equivalent. The genetic variation within the group of clones then collapses to the genotype of the single mutant clone. …A new species arises when a mutation gives an individual the ability to invade a new ecological niche, rendering it and its descendants immune from extinction during episodes of periodic selection.” (page 51) Interesting stuff!

      2. Particularly for bacteria and archaea, where a species’ genotype is more like a pool of genes that are variously distributed amongst individuals.

  8. If you randomly place n markers on a “bingo card” you will find there is a “grey zone” where suddenly the probability of a bingo goes from near 0 to near 1. Perhaps it’s something like that with the neutral drift? There must be models that look at this.

    1. I was thinking something similar. This paper doesn’t tell us much about common mechanisms of speciation, but it does say something about the size of the “bingo card” of reproductive specificity, and suggests that the “card” occupies roughly the same proportion of the genome in diverse taxa.

  9. Still very interesting, even in light of the caveats that you describe. But this is how science should work. We have some results, and that should stimulate further work by others to test its mettle against new data.

  10. Can speciation occur quicker when the environment suddenly changes or a new niche is created? Regarding a new niche, I’m thinking about the mosquitoes in the London Underground which must have evolved in a very short time.

    1. Yes; it’s manifestly clear that if reproductive barriers are the byproduct of selection (and we make a good case in Speciation that they are, and genetic drift is unimportant), then the stronger the selection differentiating populations, the faster speciation is likely to occur. Animals landing on new island, like the colonists of the Hawaiian archipelago, are probably good examplss of this.

    2. Probably you’ve chosen a bad example as the “London Underground Mosquito” isn’t unique to the London underground it seems. Or at least a lot of the necessary adaptations for ‘underground mosquito-ing’ didn’t occur in London [see below quote]…

      Taken from this Wiki:

      QUOTE: Genetic data indicate the molestus form in the London Underground appear to have a common ancestry, rather than the population at each station being related to the nearest above-ground population. Byrne and Nichols’ working hypothesis was that adaptation to the underground environment had occurred locally in London once only – many hurdles must be overcome to become adapted to the subterranean environment, and understandably it would occur rarely. This hypothesis implies that local adaptation would be expected in different locations around Europe and beyond, as each local population evolved an offshoot that overcame the problems of living underground.

      However, more recently collected genetic evidence reported by Fonseca and others suggests a single C. p. molestus form has spread throughout Europe and beyond, since populations over a large area share a common genetic heritage. These widely separated populations are distinguished by very minor genetic differences, which suggest the underground form developed recently; a single mtDNA difference is shared among the underground populations of ten Russian cities and a single fixed microsatellite difference occurs in populations spanning Europe, Japan, Australia, the Middle East, and Atlantic islands. This worldwide spread might have occurred after the last glaciations or be even more recent, due to the insects hitchhiking on world trade routes; one possibility is the international second-hand tire trade. The tires retain water in which the larvae can survive, and completely removing water from an old tire can be difficult.

  11. Very interesting at my limited level of understanding. At least I now appreciate a lot more what is meant by the term “taxa” – I’ve had it wrong for years!

  12. I imagine that during the colonization of the sterile earth there would have been rapid speciation to do with occupying the plethora of niches available and that the mechanisms changed once the niches were filled. I suppose it’s difficult to know that or determine how it worked then, though.

    I’ve always been curious about how evolution and speciation change over deep time. Why did humans appear now and succeed, or were there actually many other such species in the past who disappeared before making their mark?

    1. @ Curt Quote: “…Why did humans appear now and succeed, or were there actually many other such species in the past who disappeared before making their mark?”

      Define “succeed” – all the species alive today have “succeeded” & many species with the bad luck to go extinct were “succeeding” for a longer span of years than there’s been primates knocking about.

      Ant & beetle species have been succeeding for a lot longer than us & they’ll still be here succeeding long after we’re dust [the unfortunate truth]

      1. Succeed means that in a billion years it will still be clear who were were and what we did. Also that we are all over the world and there are billions of us.

    2. A thought about the level of variety of organisms following the abiogenesis event [or multiple such events some of them maybe simultaneous in different Earth locales] on the early Earth:

      In the era pre-single cell & early single cell there was probably no sexual reproduction** & the concept of species isn’t too useful among simple organisms when [& still today among the smallest critters] it’s all about horizontal gene transfer as much as inheritance down the generations…

      ** I have no idea when sexual reproduction began, but I assume it was after the invention of the cell wall. A long time after. If someone knows more please shoot me down in flames 🙂

      1. Jerry makes a good case in some post that prokaryote horizontal gene transfer (HGT) isn’t sexual reproduction.

        That makes sense since eukaryote sex is late – at most 2 billion years old vs 4 billion years of evolution – and hence difficult in some sense. (It is more complicated vs several linear chromosomes and the meiosis cell cycle, for two.)

        I doubt there is too much HGT going on to fuzzy up vertical lineages much, seeing how we could characterize the latest universal common ancestor based on tree studies [ ].

        You can now argue that there could have been some 100 Myrs of pre-DNA evolution with less tendency of “clumping”. But then we would need some extraordinary evidence for that.

  13. This is a fascinating topic. Would it be the case that geographic barriers are fewer for marine animals than for terrestrial, and if so does this have implications for speciation at sea versus on land? There seem to be many species of whales.

    1. “Would it be the case that geographic barriers are fewer for marine animals than for terrestrial, . . .

      I’m not so sure that is so. There are many distinct types of environments in the oceans that equate to geographic barriers on land. For one example, light, salinity and oxygen content all vary by depth and frequently there are distinct layers. Species vary by layer.

      A good example is a paper Jerry did a post on a year or so ago on the Opa. It had evolved features which allowed it to maintain a higher body temperature than most other fish in its environment allowing it to make unusually deep dives into an environment that most fish could not survive in and thus feed where there is less competition for food.

      Other examples are desert areas where nutrients are sparse and fecund areas like coral reefs.

      1. Yes, but there is no pre-speciation barrier. Suppose an ancestral fish population can swim freely at different layers. What isolates the sub-population that swims in layer A from the identical sub-population that swims in layer B before speciation? I can see that after the population speciates there could be an ecological barrier.

        1. The population needn’t be homogeneous before speciation. If there’s variation that causes some individuals to prefer layer A and others to prefer layer B, then the A-lovers are more likely to meet and mate with other A-lovers, and B-lovers with B-lovers. Speciation follows from that segregation; it doesn’t precede it.

          1. Okay, That seems reasonable. But isn’t it then more an ecological preference, rather than a geographic barrier, that separates the sub-populations and leads to speciation?

            1. I’m not sure there’s a clear distinction. If a herd of deer balks at crossing a river even though they might be physically capable of it, I think we’d still consider that a geographic barrier.

              1. I think you’re right in what you are envisioning but the ‘deer baulking river’ scenario might not be the best illustration simply because the most common conceptualization involving that scenario has the direction of selection crossing the river (i.e. a seasonal migration). It’s not that your point cannot be made using that, but it’s a case of a pre-existing reasoning channel is quite hard to resist, so fewer people will follow you. I was very confused..thought I was supposed to jump into the river and cross. I got soaking wet and an alligator ate me and I resent that.

        2. “Yes, but there is no pre-speciation barrier.

          It seems plausible to me that there is. For an example, suppose a sub-population of a species that lives in the shallows around a group of volcanic islands is separated from the main population by a storm and ends up finding a different group of volcanic islands around which they are able to survive. The two groups of islands are in deep water which the fish normally can not or do not cross.

    2. I would add the structure of ocean streams, from having studied the Tara Ocean project 2015 aggregated study on viruses (for a literature project) [“Patterns and ecological drivers of
      ocean viral communities”. Brum JR et al. Science. 2015.]

      The paper tests the marine viral “seed bank” hypothesis to some extent, finding that large local populations of microbial hosts are influenced by environmental conditions, so the microbes are, so the viruses are. These dominant communities form the “bank”, and low diversity is seen during drift along currents that can disperse the viral “seed”. Besides the paper genetic diversity data (on identified genomes or protein clusters), I noticed that the TO bacterial diversity paper in the same journal edition had similar diversity figures for the dominant viral host form (of the smallest cells) and the viruses.

  14. Great post and great intro ! The controversy, if any, is not obvious however. The authors of the paper studied multiple pairs of diploid animals, so, talking about “polyploid speciation” here is out of topic.
    The message of the article in PLOS is not : “speciation HAS TO take a very long time”, but it’s more: “barriers between species can evolve very fast, but most of the time, they are not strong enough to imped introgression all over the genome”, and they quantified the range of divergence where barriers are semi-permeable.
    After that, I don’t know how relevant is the choice of surveyed species in this study, but, it’s easy to imagine that if someone does the same study by focusing only on species which are good candidates for “sympatric speciation”, so the picture will be different. Similarly by focusing only on candidates for “polyploid speciation”, etc …
    So, the question is: what are the relative importance of “instant or rapid speciation” in the global picture of current species diversity? How much the authors are wrong when they write semi-permeable barriers are frequent in nature?

  15. “The Big Deal about the paper, however, is the fairly narrow transitional “gray zone”: the bar that marks the transition between fairly free gene flow (populations or incipient species) and full species. That gray bar extends from 0.5% to 2% sequence difference at “neutral” sites.”

    From 0.5% to 2% is not narrow, it’s very large ! Human – Gorilla = 2% Jerry !
    It seems narrow on a log-scale, but it isn’t. Taxas more divergent than 2% within a genus are not very frequent.

    1. I noticed the range too, and saw that the current estimate Human – Chimp is 1 % (1.3-1.4 % IIRC). Explaining why Homo is such a mingled set of gene flows, perhaps.

  16. Thanks for this opinion. One comment on sampling size:

    In 2009 I started a comparative population genomic project in animals, during which the transcriptome of ~500 individuals from > 100 distinct species was sequenced. The idea was that most of the molecular evolutionary literature in animals at that time was dominated by humans and drosophila, which I thought were not enough representative of the diversity of animals. This project resulted in a couple of publications that offered a comparative perspective and, I think, new insights into some aspects of molecular evolution and population genomics. These include Romiguier et al. 2014 Nature, Galtier 2016 PLoS Genetics, and Roux et al 2016 PLoS Biol.

    Admittedly, “100 species” is different from “all species”, but it is also very different from “2 species”. Of course there might be exceptions. But if we’re not qualified to try and make general statements based on this data set, who is?

    I have faced many sorts of criticisms regarding the sampling strategy in this project – “comparing oranges and apples”, “not enough individual per species”, “why transcriptomes instead of genomes”, “target tissues differ among species” – but I has not anticipated one could blame us for not including Drosophila! The whole project was meant to be everything but Drosophila. I’m sick of Drosophila. I feel like I know everything I need to know, and more, about Drosophila. Time to learn from the non-Drosophila part of the living word.


    Nicolas Galtier

    PS. To whoever might be interested: there are actually birds in this data set (tits, wrens, ducks, penguins)

    1. I’m sorry, but I think you’re being overly petulant here. My crack about Drosophila was a joke, as I hoped was clear from the exclamation mark after it.

      And seriously?:

      The whole project was meant to be everything but Drosophila. I’m sick of Drosophila. I feel like I know everything I need to know, and more, about Drosophila. Time to learn from the non-Drosophila part of the living word.

      First of all, fruit flies are part of the living world and if there were data on them, they should have been included. But your statement about being sick of flies and know all you want to know about them belies some kind of, well, I don’t know.

      The sample size should be increased and expanded, and I presume you agree with me. That was not my major quibble with the study.

      1. Honestly Jerry, you are used to share some straight opinions about many different topics, it also “belies some kind of, well, I don’t know” … don’t you think?.
        It’s a process of intent you did and people expect more from you.

        I agree that a huge amount of time, money and effort was already spent on Drosophila. I personaly started a PhD in evolution not for being “doctor in fruit flies”… We know its genome, its embryology, its ecology, its phylogeny, its demographic histories, the transposons, etc …

        What remains exciting to discover specific to Drosophila?

        1. Wouldn’t the fact that there is so much data on Drosophila be a very good reason to include it in a comparative study like this? It seems to me that if wishes could come true you’d be wishing that you had as much data for all species included in the study as there is on Drosophila. That could only enhance the study.

          And the study isn’t about discovering things specific to the individual species in the study is it? It’s a comparative study intended to find patterns in a more general context than species, isn’t it? If that is the case then, again, more data on, and a better understanding of the individual species will only enhance the study.

          1. OK I see, now I better understand the point made by Jerry, thanks.
            It’s just surprising to see that there is no such critics for 60 independent studies using only one different pair of species, but this kind of critics occur when 60 pairs are studied. It seems easy now to analyse Drosophila data like the authors did, and to add the point to this figure.

        2. No, these data do not exist for most of the thousands of species that you are referencing. And we absolutely do not have a respectable phylogeny for the genus, although multiple groups are working on it.

        1. lol, no, I let them living quietly around the trash in my kitchen 🙂

          It’s true that studies on model species are crucial for many points, but, I can be irritated by people who are not able to recognize the values of studies made from “non-model species”. In addition, I find laughable to read “(almost) everything is in my book” while there was so many advances since it was written 13 years ago.

  17. Something that should have been ironed out by now, is the legacy of theory formation in biology typically by biologists in reliance on inaccurate information – which exist as common memes among scientists – about a different field to their own – hybridization.
    One can look at the natural world and reasonably conclude that there must be major barriers to hybridization. But the basis for that, if it is better known or known to be different than, what is popularly thought among scientists, then it’s something that should be taken seriously, at least to the extent of some kind of basic awareness about assumptions built into concepts that are known not to be correct, but which are kept in place because they are useful/convenient.

    Just so that people continue to have the basic minimum of working concepts that they retain some native visibility of the gaps or potential gaps beneath the label. Might need it oneday.

  18. Besides the sampling issue:

    You express two interesting caveats: (i) what we measure is molecular divergence, not time; (ii) the molecular clock ticks at a different pace in distinct groups. Taken separately, these might appear as weaknesses of our study. I see their combination as a strength.

    The fact that, in taxa evolving at such different rates, the occurrence of gene flow can be predicted from neutral molecular divergence indeed suggests that time is not what matters in the first place. Given the
    irregular molecular clock, divergence time must differ substantially among species pairs in the grey zone. This suggests to me that the rate of accumulation of species barriers is primarily governed by the rate of
    accumulation of substitutions in general: once your genomes have sufficiently diverged, there must be incompatibilities that will prevent gene flow after secondary contact. This process seems to dominate over details regarding ecological differences and local adaptation in early diverging populations. This is what at least I was meaning in the (admittedly too vague) sentence you quote about speciation operating “in a more or less similar fashion in distinct taxa”. Populations might initially split for many distinct reasons, but why and how they eventually become fully isolated seems pretty boring and general – their genomes have diverged enough.

    Of course, polyploids might be exceptions. As far as cichlids having diverged 5000 years ago are concerned, I wouldn’t bet that they form good species (in the sense of no possible gene flow).

  19. I’m not sure why you’re focusing so much on time in your interpretation of this paper. It’s not about time at all. It’s about pairwise gene flow and genetic differentiation, neither of which is linear or even monotonic with time. The X-axis of the figure you reproduced from Roux et al’s paper is most certainly not equivalent “time since divergence.” Considering the secondary contact model they included in their analysis should make that pretty obvious, but even in the other models pairwise diversity should not be interpreted as a stand-in for time. It’s not and the authors don’t treat it that way. You’re right that comparing gene flow and pairwise diversity is not circular in this case, but as a result absolute time cannot be reconstructed.

    The real limitation of the paper, as with most ABC analyses, is that the demographic models are too unrealistic. It’s really great that they’ve included heterogeneous introgression rates and gene pool sizes across the genome, but without also allowing them to be heterogeneous over time, the results are pretty tenuous. We know that migration rates as well as real and effective population sizes and other ecological and demographic parameters all fluctuate in natural populations over evolutionary time, so any analysis that pretends they don’t won’t hold water.

  20. So I find this figure (figure 3) tricky to interpret. I don’t think that Da should be thought of as a divergence time (or species split time). Da is the fraction of sites with fixed differences between the samples. In the presence of gene flow alleles are prevented from becoming fixed differences between the species. Therefore, a pair of species that recently separated (with no subsequent gene flow) may have the same Da as a pair of species that began to accumulate RI factors long in the past but who continued to occasionally exchange alleles. As such I don’t think that the graph contradicts the idea that speciation may proceed at very different rates.

    1. Totally agree that Da is not time. I think what you’re expressing is that diverging populations can also move from bottom-right to top-left, not only top-left to bottom-right, in this figure, which I believe is correct. So a given Da and a given Prob(gene flow) can indeed correspond to very different split times and divergence histories, depending on how long the pair has moved back and forth along the continuum.

      My conclusion would be that, if one wants to qualify (or quantify) the « stage of speciation » of a given pair of populations, the appropriate unit is Da, not time. The probability that two diverging pops will eventually become two true species is apparently much better predicted by Da than by divergence time.

      The time it takes for one species to become two is variable, but the amount of substitutions it takes seems pretty constant – so we have our new speciation clock, don’t we?


Leave a Reply