New paper claims that gentoo penguins are now four species, not one

November 9, 2020 • 11:00 am

The gentoo penguin was formerly a single species, Pygoscelis papua, with a circum-Antarctic distribution. Previously, a few subspecies were named on the basis of slight morphological differences between populations inhabiting different places (they live not only on the Antarctic peninsula, but also on the South Shetland and Falkland Islands, South Georgia, and other places). They’re philopatric (breeding in the same place year to year), so the geographically isolated populations would be expected to be differentiated from one another due to natural selection and genetic drift operating on populations that don’t exchange genes with others.

A new paper in Ecology and Evolution (click on screenshot below, pdf here, and reference at bottom), suggests that this differentiation has proceeded to the point that there are now not just one gentoo species, but four.  This is a subjective decision based on both size differences and genetic divergence, but more on that shortly.  (The BBC also has a popular account of the research.)

Here are two pictures I took of Gentoos on the Antarctic Peninsula in the fall of 2019. They are notable for their white head stripe (said to look like a turban, which may be a source of their name), and their reddish-orange bills.

A colony, most of them prone against the icy wind:

And here’s the distribution from the paper above; penguin DNA and skins were sampled from the four general areas with the colored triangles (six locations total).

(From paper): Geographic range of gentoo penguins. Gray zones show existing gentoo penguin colonies while colored triangles show populations included in this study. FALK, Falklands; SGI, South Georgia Island; SSHWAP, South Shetland Islands & Western Antarctic Peninsula; and KERG, Kerguelen

First, though, you need a criterion for delimiting different species before you start making decisions about species numbers. The classic criterion is known as the Biological Species Concept (BSC), which says that if two populations maintain reproductive isolation from one another (very limited or no gene exchange) where they coexist in nature, those populations can be considered separate species. The reproductive barriers must be based on genetic differences between the populations, and can involve many different impediments to gene exchange, including unwillingness to mate, sterility or inviability of hybrids, breeding at different times of year, preferring different microenvironments within the same general area, and so on.

The BSC is obviously be best applied where populations are sympatric—that is, that live in the same place so we can actually determine whether the barriers to gene exchange are effective. But it can still work if populations live in different places BUT, when you forcibly mate them in the lab or in zoos, the hybrids are sterile or inviable. Those hybrid problems would certainly occur in nature, too, so it’s a valid inference that hybrid problems of this sort found for allopatric populations (living in different areas) mean that those populations are different biological species.

However, the converse isn’t true: if populations can produce fertile and viable hybrids in captivity, that doesn’t mean they are the same species, for they could produce some fertile and viable hybrids in zoos but not mate with each other in the wild because of different breeding periods or sexual distaste. (This is true of lions and tigers, which can occasionally produce fertile hybrids in zoos but have never been found to hybridize where they co-occur in nature.)

Thus, using the BSC on allopatric populations is a subjective matter unless their hybrids are sterile or inviable.  And this uncertainty holds for gentoo penguins. The populations samples in this study come from geographically isolated populations, and I don’t know anything about inter-population hybridization in zoos. One can’t, therefore, use the BSC on gentoos, a concept that I consider asthe best species concept for a number of reasons. (For a discussion on why I favor the BSC, see Chapter 1 and the Appendix of my book with Allen Orr, Speciation.)

The authors, then, default to another species concept when sampling the gentoos: the phylogenetic species concept (PSC). It’s defined as “the smallest diagnosable cluster of individual organisms within which there is a pattern of ancestry and descent.”

I discuss the PSC in the appendix of Speciation; it has several problems that make it less than optimal.  “Diagnosable” means that the clusters can be told apart. But this doesn’t necessarily reflect genetic differences between populations. Suppose that one population of penguins lives on an island surrounded by ample marine food, while the other population has less food. The former population could produce large penguins, while the other population, with meager food, could comprise small penguins. You might be able to tell the adult penguins apart by size, making the two clusters “diagnosable”, but it wouldn’t be a meaningful difference, not reflecting anything about the genetic distinctness of the population. Such differences could occur in just a single generation due not to evolution but to effects of the environment on development.

Further, if populations are geographically isolated, as with those on different islands, then there will be a parental pattern of ancestry and descent for each island. Each individual will be born to parents from its own island, so that criterion is met. Couple that geographic isolation with “diagnosable” traits, even nongenetic ones, and you get a new species. But what does that mean?  Very little, at least in terms of biology. As I’ve emphasized, “species” exist to describe the “lumpiness” of nature in one location: the fact that, say, birds in one forest don’t blend imperceptibly into one another, but fall into distinct groups. The groupings reflect lack of interbreeding in nature, and any sensible species concept has to deal with that lumpiness in one area. The origin of species is the same as the origin of distinct “lumps” of plants and animals that can be told apart in a single place.

So if you move a population of Asian humans to one island and Caucasian humans to another, and force them to stay there forever, the PSC automatically makes them different species of humans, for each individual can (by both looks and genes) be diagnosed, and the populations maintain a pattern of ancestry and descent. But humans are known to be a single one species with many different morphologies and statistical differences in gene frequencies from place to place. Asians aren’t one species and Caucasians another, as they readily interbreed where they coexist. So what does it meant to call the Asians on island X as a species different from the Caucasians on island Y? Nothing.

I went into this matter because, to some extent, this is the situation with the gentoo penguins. They live in different places, and while there are diagnosable genetic differences between them, just as there are for human populations if you use a combination of many genes, and while there are also morphological differences between the gentoo populations, those differences are purely in size and shape.  In the end, if you think the DNA differences between the gentoo populations are sufficient to make them different species, that’s purely a matter of taste, for there is no obvious correlation between DNA differences and ability to produce fertile hybrids were the populations to co-occur in nature. And we don’t know whether the morphological differences between the gentoo populations are even based on genetic differences, as they could be developmental differences induced by the different environments that, as the authors admit, exist in the different populations of gentoos.

That said, here are the data.

For morphology, the authors looked at 39 penguin skins from five locations collected in two museums. Measurements were taken on nine linear characters involving beak and body size. Note that every character is related to body size. (See the paper for the list of traits).

For DNA, the authors used 69 DNA samples from five of the six populations described in the present study. They used more than 10,000 DNA sites for their analysis of genetic distance (separation), so it’s a very large sample.  This sample and some of the analysis described in this paper comes from a 2018 paper in Molecular Ecology by Clucas et al.

The results can be stated briefly (I hope!)

Morphology.   The penguins generally sorted out by location (four clusters) based on statistical combinations of the body-size traits (principal components analysis and linear discriminant analysis), but many of the individual penguins were NOT morphologically diagnosable by even combinations of their size measurements. You can see that from the overlap of the clusters in the two plots below (each color of a dot is one individual from a given population; triangles are population means). Here are the scatterplots for size traits:

(From paper): Principal Components Analysis of linear morphometrics. BH, bill height; BWB, bill width at base; BWG, bill width at gonys angle; CL, culmen length; ML, manus length; MTL, middle toe length; RL, radius length; TML, tarsus length


(From paper): Linear Discriminant Analysis of linear morphometrics. Circles represent individual specimens with triangles showing the lineage mean

The lack of diagnosability can be seen because the position of a dot does not tell you with 100% certainty its color; that is, you can’t diagnose an individual with absolute certainty even from statistical combinations of the nine body size traits. In general, populations are differentiated, but they don’t fall into completely separated clusters, which is required for “diagnosability”.  The authors also note that, apart from size, there were no other obvious traits in plumage that could be used to tell the populations apart. So we have populations somewhat differentiated by body-size traits (and thus shape traits), but that’s also true for many species that are not “split” into multiple species, including H. sapiens.

DNA.  The DNA was much better at diagnosing individuals. In fact, using the 10,000+ DNA sites, an individual could absolutely be assigned to population, which means there’s been pretty substantial genetic divergence between the four localities (Falkland Is., Kerguelen Islands, South Georgia Island, and South Shetlands/Antarctic Peninsula). The clustering based on DNA shows absolutely distinct clumps, and putting the DNA into a phylogeny (family tree), one sees that the four groups form a distinct pattern of ancestry.

(From paper): Principal Components Analysis based on genetic data. The amount of variance explained by each principal component (PC) is displayed on the inset bar graphs and on the axes, and the number of PCs retained is indicated in black

The phylogeny of the four groups (the “100” means that the clusters are extremely strongly supported using “bootstrap” analysis:

(From the paper): Best Scoring maximum likelihood phylogeny based on 10,108 neutral SNPs. Support values shown for branches that received >90% bootstrap support

How many species?  Based on these data, the authors divide gentoos into four species in the genus Pygoscelis:

P. papua in the Falkland Islands
P. ellsworthi in the South Shetland Islands
P. taeniata in the Kerguelen Islands, and
P. poncentii from South Georgia Island.

The upshot:

The four populations have differentiated fairly strongly in their DNA, which means (since the DNA sites were “neutral” ones, reflecting the passage of time), these populations have been geographically isolated from each other for a long while, though I can’t say how long. The populations are diagnosable from DNA and apparently have a pattern of ancestry and descent, so they fit the concept of phylogenetic species.  But if you use morphology by itself, these are not phylogenetic species, as they aren’t diagnosable from morphology. But all you need is one or a few traits to diagnose a population as a new species according to the PSC, so these are phylogenetic species.  I would be wary of diagnosing them from morphology, anyway, since these are size traits and likely to be affected by local environments as well as genes.

But are these BIOLOGICAL species? The answer is that we don’t know.  I suspect that, if a penguin from Kerguelen were to somehow find its way to the South Shetlands, it could mate with the locals and produce fertile hybrids, but that’s just my feeling. Because this doesn’t happen, and I don’t know about the situation in zoos, we can say only that we’re not sure whether these are biological species. One could make a somewhat informed judgment by seeing if other species of birds with similar genetic divergence would be unable to produce hybrids where they co-exist; that would give us some idea of whether this amount of genetic divergence between the penguin populations would prevent the formation of fertile hybrids. But of course each group of birds will have its own “speciation clock” (for one thing, the generation times differ), so that would also be a guess, though an informed guess.

Should the authors have named these as different species? My answer (assuming I adhere to the BSC) is, “I don’t know; it depends on what we can find out from their reproductive compatibility in captivity.” If they can’t produce fertile and viable hybrids in zoos, or simply refuse to mate with each other there, that gives us a pretty good idea that they’d be reproductively isolated were they to co-occur in nature; hence they could with some confidence be called different biological species.  But PSC species don’t always coincide with biological species, so I reserve judgment and just say that the authors are “splitters”, looking for reasons to divide up populations into distinct species.

One reason biologists like to do this is because it’s easier to conserve animals that belong to different species than those that simply belong to different populations of a single species (subspecies are, under America’s Endangered Species Act, also conserved by law). Splitting thus enables you to try to save more animals, as is evident from the tweet below.  Fortunately, gentoos aren’t endangered, though some of the sampled populations have decreased in size over the last two decades. And, if global warming hits Antarctica and the southern oceans hard, as it is likely to do, eventually all the penguins will go extinct, and splitting will have made no difference.

Note the implication in this tweet that the species were “identified,” as if there are pre-existing and real entities out there that have finally been pinned down. But of course that’s not true; these populations differentiated genetically slowly over time, and there’s no one instant where a new “species” has popped into being and can be identified as such.

h/t: Paul


Tyler, JBonfitto, MTClucas, GVReddy, SYounger, JLMorphometric and genetic evidence for four species of gentoo penguinEcol Evol2020001– 11

30 thoughts on “New paper claims that gentoo penguins are now four species, not one

  1. Very good discussion of species concepts (of course!)

    One could just as easily group these into just two species. Even though it is not definitive, I would like to have seen how th degree of genetic differentiation between “species” compares to that of closely related universally-accepted penguin species. If the branch lengths for these new “species” are as deep or deeper than the lengths of recognized species, I would regard that as giving some support to the splits. This criterion might also distinguish between the option of two versus four species.

    Maybe the authors covered this?

  2. I’ve often wondered if there was any objective criterion for separating species based just upon DNA differences. There seems to be a lot of that around these days.

      1. Thank you. Thought so.
        This question has been coming up a lot with regards to a genus I’m currently working on.

  3. Yea science post!

    Are there are zoos with gentoo penguins whose ancestors came from two or more of these places? If we measured their genetics, we could quickly see if genetic mixing has already occurred?

    I see that they also are found on a number of other islands, so that adds a requirement of good record keeping as to penguin provenance. But also gives an opportunity to see if all of these colonies are genetically isolated from each other. Seems like a nice way to potentially speciate!

  4. Really liked this post. Not least because I found a word that was new to me :

    Now I just have to engineer an opportunity to drop it into conversation…

  5. If there were some clear, even if small, morphological differences between these populations, I’d be happier calling them species. To me, it seems more like they’re in that long transition state where they may become different species, but maybe not.

    However, I do understand naming things as species to get them protection. (In the U.S., subtaxa of plants can’t be treated as legally endangered species.) In one ambiguous case, I chose the species level mainly because of the potential legal protection. In another, though, I had the unhappy job of telling the client that the genetic evidence available didn’t support separation.

    1. I often have the same dilmemma. I think in general we need to try to ignore these consequences of our decisions, and just do the most consistent and honest job we can. Otherwise we end up cheapening the whole concept of species and make it harder to convince people to preserve real species.

      1. Indeed. If we use a political motivation to drive scientific conclusions then perhaps we are using a Critical Species Theory?

  6. Good science. We know more than we did, which hopefully will help us with these birds, even if it doesn’t help us pigeonhole (heh) them into neat categories.

    I guess the lumper answer to the splitters would be: over my gentoochas.

  7. While reading this post, I kept asking myself, what is the main reason biologists want to split these species apart? Then I finally came to your last paragraphs and viola. Makes sense. Thanks for this interesting post. I also enjoyed seeing again the gentoo photos you took on your Antarctic voyage. Seems like years ago. Everything pre-pandemic seems a lot further away.

  8. Hoping you’d post on this, thanks for that.

    Lumpers v Splitters will never die……

    Nice example of LDA too, would make a change from irises if the data are available.

  9. I appreciate this and the informative discussion, since I haven’t had time to look into the paper (and given the lack of comparison with other species I now doubt I will)!

    What did catch my interest is that the context of penguin diversification – which source locale of the New Zealandia plate has mostly submerged – once again is plate tectonics. I wondered what made Kerguelen (and Heard & MacDonald Islands) sit so far from any continental plate, and remembered a fresh paper on them being the products of an extreme midoceanic ridge system:

    Kerguelen Plateau Is Earth’s Longest Continuously Erupting Supervolcano

    The Kerguelen Plateau in the Indian Ocean is the world’s longest continuously erupting supervolcano dating back into the Cretaceous, according to research published in the journal Geology. This discovery is quite exceptional, considering that these types of eruptions typically last just 1-5 million years.

    For over 30 million years, from around 122 million years ago to 90 million years ago, basaltic lava erupted from fissures on the seafloor building up a volcanic plateau, eventually breaking sea level.

    The study suggests that the Kerguelen Plateau saw such a long and steady run of volcanic activity because of its unique configuration: a mantle plume combining with an active mid-ocean ridge. The Indian Ocean is crossed by three mid-ocean-ridges, the outer limits of three large tectonic plates, including the African, the Antarctic and Australian plate. Until a few million years ago, the Kerguelen Plateau was located exactly above a hot-spot sitting along the mid-ocean ridge between the Antarctic and Australian plates.

    [ ]

  10. It should be noted that the multivariate analyses they use for the morphological differences are statistically invalid. Such analyses require that the input variables (here the morphological data) are independent and as they are mostly a function of size, they clearly violate that assumption. Theoretical work has shown that using such data will overestimate the difference between samples. This means that the found morphological differences are likely even less pronounced than in reality.

    It is also a pity the paper ignores completely the evolutionary history and potential gene flow between populations. PSC papers tend to do this, mayb because it complicates the picture considerably in most cases? PSC is an interesting concept, but just like the BSC it is a discrete concept applied on a continuous scale, so any concept on it’s own is bound to not work in many cases.

    RE Gentoo Penguins in zoos. At least in Europe both papua and ellsworthi are kept and in recent years also in mixed colonies (though the nominate is far more common), I don’t know about any mixed pairs and breeding attempts. One should contact Jo Elliot from Edinburgh Zoo for that, who is the studbook keeper.

    1. Can you clarify exactly what you mean here, in particular where you say “independent”?

      PCA is certainly size-dependent though, which is why people often standardise the variables to unit variance (aka using the correlation matrix).

      1. Realised what you probably meant just after posting the above, and I should have said “scale dependent” there.

        The morphology PCA picture clearly shows the usual pattern of PC1 being basically “size”, i.e. all the morphological variables are +vely correlated (all arrows point left) so larger individuals plot on the left and smaller ones on the right of that picture.

        1. What I meant is the the input variables are not independent, as they are basically all measurements of size. That is indeed what the results also show. They should have done a stepwise selection to reduce the number of size parameters, as with many non-independent variables the analysis will actually overestimate the differences between the penguin individuals (which are small to start with). So I think we are actually on the same page

Leave a Reply