The orders of modern placental mammals originated after the extinction of the dinosaurs

February 9, 2013 • 2:46 pm

by Greg Mayer (Updates below.)

A new study just published in Science by Maureen O’Leary and colleagues examines the phylogeny of 40 fossil and 46 extant mammals based on a very large data set of morphological and molecular characters (the latter only from the living taxa). The study has gotten a fair amount of attention in the press, where it seems to have been misinterpreted; more on that later. First, let’s see what they were trying to do and what they found.

There are three major groups of mammals alive today: the egg-laying monotremes (the platypus and the echidnas), the marsupials (opossums, kangaroos, bandicoots, etc.: a few hundred species) and the placentals (cats, dogs, cattle, deer, and all the rest, including primates: several thousand species in about 18 orders in all). The placentals are the overwhelming majority of extant mammals, and dominate the mammalian fauna of all parts of the world except Australia, which has mostly marsupials. (There are a fair number of marsupials in South and Central America, but they are still outnumbered by placentals.)

In the fossil record, although basal placentals are known from the Cretaceous (some of these fossils are disputed, including by O’Leary et al., but all agree there were some), the great radiation of placental mammals did not occur until the early Cenozoic, after the extinction of the dinosaurs (at least those that had not evolved into birds) at the end of the Cretaceous, about 65 mya. Although the first two-thirds of mammalian history had occurred during the Mesozoic (the “Age of Reptiles”), they really broke out, biodiversity-wise, in the Cenozoic (the “Age of Mammals”).

There’s been considerable debate over whether the placental lineages that radiated in the Cenozoic arose just before the radiation (the “explosive model”), had existed since earlier in the Cretaceous but did not radiate until the Cenozoic (the “long-fuse model“), or had undergone considerable diversification in the Cretaceous (the “short-fuse model“). The latter model would require that the fossil record be seriously incomplete, but has been supported by various molecular phylogenetic studies that estimate various splits among the extant placentals to have occurred well before 65 mya.

Models of placental mammal radiation (Rose, 2006).
Models of placental mammal radiation (Rose, 2006).

The chief question O’Leary and colleagues addressed was which of these models is correct. To do so, they scored over 4000 morphological characters (including soft-tissue characters generally not scorable in fossils) and utilized 27 nuclear gene sequences to estimate the branching sequence. They then added in the known stratigraphic range of the fossils to get a phylogenetic tree (their Fig. 1) that looks very much like the explosive model above, except that the common ancestor of extant placentals (the “P” in the figure) occurred jut over the line, in the earliest Cenozoic rather than the Cretaceous, making it even a little bit more explosive-y. “Explosive it is, sir!”, as Apu on the Simpsons might have put it.

They make two further interesting inferences from their tree. First, they use their morphological data set to estimate what this earliest Cenozoic common ancestor of all placental mammals looked like. It looks like this:

Hypothetical early placental mammal (O'Leary et al., 2013).
Hypothetical early placental mammal (O’Leary et al., 2013).

With such a large data set this is interesting, but it does look pretty much like what people have long thought early placentals would look like. Remember, this is a hypothetical common ancestor, not a newly found fossil.

Second, because much of the breakup of the Mesozoic super-continents had occurred by the early Cenozoic, they infer that a lot of dispersal occurred in the placental radiation, and not just passive floating around on the drifting continental plates. The exact arrangement of lineages in the tree is also of interest, and will be discussed and debated by mammalogists. It’s not clear to me that a huge data set is necessarily an advantage in inferring this large scale phylogeny, because we don’t understand the dynamics of conservatism and lability of morphological characters in the way we do for genetic sequence data. Our understanding of the latter allows us to select genes and use methods of analysis appropriate for a particular question. Using thousands of morphological characters seems a bit too reminiscent of the old pheneticists’ hope that if they could score enough characters, “parametric overall similarity” could be known (phenetics didn’t pan out as hoped). I hasten to add that morphological characters are more difficult because they are more complex and more crucial to the organism, and consequently more interesting– indeed, what most biologists are really interested in– not because there is something wrong with studying morphology.

In the media, there has been considerable confusion about this study, in part because the distinction between mammals and placental mammals has not always been kept clear. The  New York Times initially led with the headline:

“Common Ancestor of Mammals Plucked from Obscurity”;

but, of course, the study is not about the common ancestor of mammals, but only placentals. And furthermore, there is no particular known fossil which is being identified as or compared to this placental common ancestor; the ancestor in the picture, as stressed here, is hypothetical. Yet, the Times article identifies Protungulatum as the ancestral placental, O’Leary et al. most definitely do not do do: they identify Protungulatum as a member of the lineage that gave rise to (most) hoofed mammals (i.e. quite far from the common ancestor of all placentals). Protungulatum is the oldest known member of the clade that includes all extant placentals, but that does not make it the common ancestor.

This misunderstanding has infected the news media, and spread widely. Gizmodo labels a picture of the hypothetical form (a version of the figure above) as Protungulatum, and states

This rat with way too many sharp teeth is your great x 4 x 10^6-grandmother. That’s what scientists have discovered after six years of research—the Protungulatum donnae is the common ancestor to all mammals, from humans to horses to lions.

This is pretty much completely wrong. And UPI labels the same figure as

An artist’s rendering of Protungulatum donnae,

which it isn’t at all. The Times, at least, subsequently changed its headline to

“Rat-Size Ancestor Said to Link Man and Beast”,

which is still pretty obscure, but not actually wrong. But the article retains its misstatements about Protungulatum (at least last I checked). The Times did correct another error. They had initially stated that only a single mammalian lineage had survived the end-Cretaceous extinction, but it is known that there were at least four surviving lineages (one monotreme, one marsupial, one placental, plus one multituberculate– a now extinct mammal group which survived the end-Cretaceous extinction, but died out in the Oligocene). The article has now been corrected to say that the study concludes only one placental mammal survived (which is indeed what its major conclusion is).

The article in the BBC was better, getting the headline right:

“Earliest placental mammal ancestor pinpointed”

and not mentioning Protungulatum at all. The BBC front page headline, however, was off:

“Earliest mammal ancestor pinpointed”.

For a mildly critical take on the study (not the news coverage) by a more molecularly oriented mammalogist, see Anne Yoder‘s Perspective in Science.

UPDATE. I had left the following comment out, because I thought I was saying it too often, but this paper really shouldn’t have been published in Science. There is much too much data, methods, analysis, and discussion left out of the paper because of Science‘s severe length limits. There are two online “supplements”, one at Science and another at The one at Science is 132 pages long. The morphobank supplements are not organized as a file, so it’s hard to tell how much is there, but it’s a lot. Now some of this material (e.g. lists of which authors examined which specimens) need not be published, but it’s simply impossible to fully understand or critique the paper with out referring to a great deal of this material, which is not readily available to someone in possession of a copy of the paper. The authors have shortchanged themselves and their readers by publishing in such a venue. I was moved to add this update after an alert reader noted an error in my statement of the number of species included, and I had to pore though the supplements to verify the correct numbers because of ambiguous wording in the paper. I was able, while doing so, to confirm that the same extant 46 species were used for the genetic and morphological analyses. (And since I’m kvetching, I’ll note that the authors substitute the grotesque, poorly defined, and unnecessary word “phenomic” for “morphological” (or a similar word) throughout their paper.)

UPDATE 2. The errors in the media coverage do not stem from SUNY Stony Brook’s press release, which correctly summarizes the claims of the paper.


O’Leary, M.A., et al. 2013. The placental mammal ancestor and the post-K-Pg radiation of  placentals. Science 339:662-667. (abstract)

Rose, K.D. 2006. The Beginning of the Age of Mammals. Johns Hopkins University Press, Baltimore. (Google Books)

Yoder, A.D. 2013. Fossils versus clocks. Science 339:656-658. (abstract)

38 thoughts on “The orders of modern placental mammals originated after the extinction of the dinosaurs

  1. The fourth major mammal lineage, the multiturberculates, also survived the K-T extinction but went completely extinct during the Oligocene, so they always get left off the list.

      1. Cool! I’ve had something of an obsession with them since I read Robert Bakker’s novel Raptor Red when I was in junior high, though sadly, the feathered dinosaur revolution (including, of course, my pseudonym) has left me unable to enjoy that book as much as I used to.

  2. This is curious and makes one wonder about previous calibrations of clock-like genes. For example, the paper by Bininda-Emonds et al. in 2007 clearly shows some placental mammal diversification before the K-T boundary, and it used a 66-gene alignment (51,000 bp) with 30 fossil calibration points. Is this new paper really more robust?
    I had also read, for example, that the most recent common ancestor of the primates, let alone all placentals, lived nearly 80 mya. And there is an early primate with the opposable toe around 56 mya (only 9 mya into the Cenozoic). I wonder if the morphological emphasis here is misleading.

    1. I have to say that I am skeptical about this “robustness” of this new paper rather than that of previous work. Particularly, their time-calibration is based on minimal fossil ages for each of their lineages. The fact that the absolute minimal ages of these clades lies this side of the KT says nothing about the real timing of splitting events on either side of the boundary. The older a fossil is, the less likely it is for it to be close to the true origination time for its lineage. It is quite unfair to assume that the first appearance of a lineage in the fossil record resembles anything like the actual origination time of that lineage. I find it particularly disingenuous that the study talks about how they are estimating minimum ages, but then interprets their estimates as if they are actual age estimates (e.g. “Ghost lineage estimates are minimum divergence dates and may underestimate the timing of actual splits.” followed by “Our results also imply that the total clade Eutheria (all species more closely related to Placentalia than to any other living species) is younger than estimated from prior studies.”).

      Studies like Bininda-Emonds et al. (2007) take this sort of problem into account by using fossil calibration points as minimum ages, but allowing splitting events to be older than their oldest fossil occurrence. I have my reservations about the methods used by Bininda-Emond et al., but most of the choices they make are motivated by a balance between the size of their data set and computational limitations.

      In sum, I think this paper adds very little to our understanding of timing of mammalian diversification.

  3. Quick correction: that’s 86 extant AND fossils specimens that were included in the study : 46 extant and 40 fossils

      1. Similar considerations as you noted in your update: the ambiguity concerning the number of species is just one of many problems that plague the correct appreciation of this paper. Reviewing this work must have been hell!

  4. I read the NY Times before I read this article. I found the NY Times article to be confusing because it interchangeably referred to mammals and placental animals, leaving me adrift regarding where the non-placental mammals enter the picture. I was torn between concluding marsupials are placental animals or there was something I was not understanding. After all, I am not a biologist, so the article is probably correct and I am the one who is mixed up. Then I read this article and now I feel that I am sane after all, it was the journalist who is crazy. Thank you for the clarification. And don’t put your trust in the NY Times.

  5. I notice that the erroneous Times piece was written by John Noble Wilford. They really should keep him away from biology articles and let Carl Zimmer do the heavy lifting on things like this!


  6. Hi Greg [GCM] Thank You

    I find it wonderful that you take the time to write such detailed, informative & lay-readable posts. As a lay reader I’m honoured that professionals such as you, Matthew & Jerry share your insights with people like me

    I read everything here on WEIT & I’m obliged to write this here & now in case you think that technical posts reduce readership [judged by number of comments]. It’s simply that people like me have nothing to add.

    More please !

    1. I echo this well-expressed sentiment!

      (You also make a darn good Ceiling Cat, Mr. Mayer… or perhaps this is a pantheistic website?)

      1. I third the motion – I don’t have anything to add or really comment on as this is not my field, however this is a good read as well as being educational, just like the other posts here. So I just wanted to say that I read the article, and Huzzah for Science 🙂

        I love the parentheticals in your post. I know parenthesising is frowned upon in some (most) circles, but it gives a proper sense of how facts are not always just black and white, and gives a sense of the provisional nature of science too.

  7. I would like to agree with Michael Fisher above, on all counts. I really appreciate your efforts in presenting the sciency stuff. I’ve only got a B.S. in zoology so I can’t comment, but love to read it.

    1. I too agree, and thanks for taking the time to make this subject “readable” by someone like me that has only limited higher ed. I enjoy reading everything on this site — getting my WEIT PhD.

  8. So I am understanding: this paper says our common, placental mammal ancestor dates to after 65 mya (the time of the death of dinosaurs), not 80 or more mya as previously thought.

    I’m not seeing how this all fits with the timeline of connected land masses and then disconnected land masses. From my look at representations of the earth, the “Old World” (Africa) was far separated from the “New World” (e.g, S. Amer.) at 65 mya. I’m not piecing together the way that ancestor mammals could have possibly dispersed to such distant land masses after 65 mya.

    Also the SUNY press release says current day elephants and others (a branch called Afrotheria) originated in the New World. So aren’t the authors implying our common ancestor also originated from the New World?

    1. That is an interesting question. The implication seems to be that the placential mammals in Africa came from the Eurasion continent when Africa hooked back up rather than being endemic to Africa.

      1. I have an excellent Ap called EarthViewer on my iPad. I don’t see “Africa … hook{ing} back up” once separated from what is now S. America at 120 mya.

        At 80 mya Africa is connected to nothing and is two land masses. Eurasia is mostly underwater. S. America is mostly underwater and connected to nothing.

        At 65 mya at least Africa is one landmass but is not connected to anything. Of Eurasia, the part that is now Europe is mostly underwater. S. America is a very recognizable shape, but it is not connected to anything.

        So the mystery (or my misunderstanding) remains.

  9. I read Donald Prothero’s Evolution: What the Fossils Say and Why it Matters, and he attributes several fossil teeth from Mesozoic sediments to current mammalian orders, implying (if these teeth are correctly identified) either a long-fuse or short-fuse rather than an explosive model. Does anyone here have any information or thoughts on this?

    Another point that occurs to me is that molecular clocks apparent date the eutherian-metatherian split to ca. 170 million years ago, which fits well with Juramaia sinensis being the oldest known eutherian at ca. 160 million years old. Is there some reason the molecular clocks would be more accurate for the origin of the metatherians than for the origin of modern placental mammals?

  10. Great post, Greg, with a wonderfully precise headline. I have been shouting at the radio and news websites for the last few days as journalists get it all hopelessly wrong, so I appreciate your analysis of media coverage.

    And you’re right about the disservice to their reserch by publishing such extensive material in a short-format journal like Science, although we both know that if any of the key authors was up for promotion or was looking for a job, that explains why. Unviersity panels are seduced by single words – Science, Nature, Cell – which are deemed as accurate and sole proxies of quality.

  11. I just read the paper last night and I agree completely with this analysis of it. I too was frustrated by the confusion in the popular press (multiple exercises in missing the point) and also wished it was a longer paper in another journal so I could read more about how they built their tree. I find the previous genetic data suggesting Cretaceous origins to be pretty convincing. I think the appropriate refutation of those ideas would be to demonstrate calibrations errors in their molecular clock estimates instead of what we got.

  12. Thank you very much for posting this. It’s a tricky subject to get one’s head around and you’ve clarified it very well.

  13. I constantly despair at science reporting.

    One minor issue that concerns me a bit about this post, however, is that its first part implies that ranks above species, such as in this case that of order, have any meaning whatsoever. There is nothing about a group of organisms that is reproducibly discoverable “orderness” or “familyness”; ranks are arbitrary. Therefore the several orders of the placentals (or the insects, etc.) are simply not comparable or countable.

  14. I am in partial disagreement with Alex SL. I think species are, most of them, real bounded entities in a slice of time. Higher categories, if monophyletic, can be argued to be just as real. However, in many cases, the monophyly is not worked out as well as we would like.

    Recognition of higher categories is not arbitrary in the sense that they are constructed without reason, and not subject to criticism. They are constructed by people like me to express our understanding of relationships the best we can. They face the same standards of description, diagnosis, and differentiation from similar groups as do species.

    Paleontologists, and modern workers in diverse and poorly known groups, have a problem in secure recognition and differentiation of species. I have looked at a lot of fish skeletons, and there are undoubted good species which cannot be differentiated by examining their skeletons. At least, I could not. Sp I am not one to criticize paleontologist for counting or discussing higher categories.

    I suppose it weakens my thesis to comment that the first new species I described is now in its fourth (and one hopes last) genus.

    1. And they are constructed by people like me, with good reason. I agree that a monophyletic group has a reproducibly discoverable biological reality: it includes all descendants of one common ancestral species.

      But it does not have a reproducibly discoverable characteristic that says “I am an order as opposed to a suborder or a class”. Supraspecific ranks are entirely arbitrary human-invented labels, and monophyletic groups at the same rank can be of very different ages, phylogenetic diversity and raw species numbers. Compare the animal families Hominidae and Coccinellidae, for example.

      Therefore a sentence like this:

      The orders of modern placental mammals originated after the extinction of the dinosaurs

      or even the always popular “X is the largest family of Y” does not actually have any scientific meaning, and thinking that supraspecific ranks have a biological reality leads to all manner of faulty conclusions.

  15. Most of the article claims are clearly not proved by anything new. As previously said, there is a real problem with the dates. The article admit ghost lineages of 60 My without any fossil found, but claim that there is no fossil gap during Cretaceous where there were obviously less lineages and less preserved fossils… how such a nonsense could be published in Science ? We clearly never found the first placental mammals, and a great matrix cannot settle the problem.

    But there is more.

    Another key claim in the abstract is the “Placental root” unraveled with the Xenarthra (species from South America) as the sister-lineage of all other placental mammals. This claim is quite surprising because the node is extremely hard to resolve with genomic data.

    I read all their supplemental data and it shows clearly that their data could say nothing about species relationships. Their “phenomic” data alone give a tree in complete disagreement with the molecular litterature (there is no well-established super-order such as Afrotheria), proving that their 4000 characters matrix is completely useless to say anything about species relationship. On the other hand, no molecular tree inference of their supplemental data give Xenarthra as sister group of all the other placental mammal (which is in agreement with all the litterature where this hypothesis is clearly marginal).

    In a nutshell, the tree of the article is the well-known phylogeny of mammals, with difficult nodes resolved with the help of phenomic characters that give obvious false relationships for 60% of the nodes. This just doesn’t make sense.

    The reconstruction of a placental ancestor with the help of morphological characters that cannot resolve species relationship is clearly weird.

    If you want another point of view about this question, see this recent publication of my lab : (complete article available here : ). Genomic data (not 26 genes…) alone give a different story about the most recent common ancestor of placental mammals.

  16. Full disclosure: Maureen is a colleague in my Department at SBU.

    Precisely because it WAS published in SCIENCE is the reason folks around the world are reading, talking and pondering its data and inferences. Yes, I also regret the massive amount of information buried in the supplemental info, typical of most papers in SCIENCE and NATURE these days, but to suggest that they should have selected another forum seems pretty far off the mark to me.

  17. I have great reservations with how divergence dates were calculated in this study. As pointed out by other commentators, the study on the one hand acknowledges that their estimates should be considered as minimum clade ages only, but on the other hand, they make bold claims such as the last common ancestor of placentals having lived after the K-Pg event. For a different take on divergence time estimation in mammals, the interested reader may consult our own work published in Proceedings of the Royal Society:

Leave a Reply