Don’t blame the dinosaurs

The mammalian tree is rooted deeply and branched early!

(click for larger image)

All orders are labelled and major lineages are coloured as follows: black, Monotremata; orange, Marsupialia; blue, Afrotheria; yellow, Xenarthra; green, Laurasiatheria; and red, Euarchontoglires. Families that were reconstructed as non-monophyletic are represented multiple times and numbered accordingly. Branch lengths are proportional to time, with the K/T boundary indicated by a black, dashed circle. The scale indicates Myr.

That’s the message of a new paper in Nature that compiled sequence data from 4,510 mammalian species (out of 4,554) to assembly that lovely diagram above. Challenging the ‘conventional wisdom’ that mammalian diversity is the product of an opportunistic radiation of species after the dinosaurs were wiped out at the end of the Cretaceous 65 million years ago, the authors instead identified two broad periods of evolutionary expansion among the mammals: an early event 100-85 million years ago when the extant orders first appeared, and a radiation of modern families in the late Eocene/Miocene. A key point is that there is no change in rates of taxon formation across the Cretaceous/Tertiary (K/T) boundary—mammalian diversity was rich before the dinosaurs disappeared.

In addition to the pretty tree, the authors plotted lineages through time of these mammalian groups—the plots pass throught the K/T boundary (the red line) without even a stutter. The second graph below shows rates of diversification, and the peaks occur well before and well after the K/T boundary.

(click for larger image)

a, Lineages-through-time plot for all (blue), placental (green) and marsupial (orange) mammals. Filled circles indicate when resolution in the phylogeny dropped below 85%. b, Net diversification rates: stepped line, rate in each age or sub-epoch; solid blue curve, rate inferred from a GAM of rate against time (χ2  =  241.5, estimated degrees of freedom  =  14.75, P << 0.001, adjusted R2  =  77.6%, deviance explained  =  20.3%); dashed curves, 95% confidence intervals. c, Counts of mammalian genera in each sub-epoch (Late Triassic to Late Eocene) according to the Unitaxon database. Red and blue lines represent genera whose families diversified predominantly before or after the Palaeocene/Eocene boundary, respectively. Throughout, the red vertical line is the K/T boundary and grey lines separate Cenozoic epochs.

Now the question is … why the delay? Why weren’t modern mammals expanding into new niches once the dinosaurs disappeared? One reason might be that the early exploiters were oddball (from our perspective, at least) groups like the archaic ungulate and multituberculates (ratty creatures, now extinct) that would have filled those places first.

Another is a matter of perspective. Big animals are the most obvious creatures in a biome to our eyes, but they are not the most important or diverse—we have a perceptual bias for the charismatic megafauna. The KT extinction was devastating for just those animals we are attuned to recognize most readily, but may not have been quite so troubling to others; the bacteria didn’t even notice, insects would have gone on buzzing (except those with dependent parasitic relationships), plants may have seen one minor class of leaf-nibblers disappear, the rat-like and lizard-like creatures of the undergrowth might well have found their environments relatively unchanged. That transient period of intense trauma might not have been as disturbing to the overall life of the planet as we big apes might like to imagine.

One other important lesson along the same lines is that mammalian diversity arose quietly and gradually among small creatures; it is again a consequence of our bias that we imagine the triumphs of mammalian evolution were the appearances of elephants and whales and tigers and of course, large apes. The important business of evolution was going on in creatures not even big enough to reach our knees, small scurriers who were founding independent dynasties 100 million years ago. We just didn’t pay much attention to them until some subset of their descendants grew large enough to trample us or eat us.

Bininda-Emonds ORP, Cardillo M, Jones KE, MacPhee RDE, Beck RMD, Grenyer R, Price SA, Vos RA, Gittleman JL, Purvis A (2007) The delayed rise of present-day mammals. Nature 446:507-511.


  1. #1 David Marjanovi?
    March 29, 2007

    That’s the message of a new paper in Nature that compiled sequence data from 4,510 mammalian species

    No, they did not.

    They took the literature, compiled a supertree from the (hopefully) phylogenetic analyses they found there, got sequences of 66 genes for as many mammals as possible, and then used those sequence data plus 30 calibration points to estimate some divergence dates on that supertree.

    That’s right: apart from its size, it’s just yet another attempt at molecular dating.

    And not a very good one. The age of the basal divergence (between the monotremes and the rest) was fixed at a rather arbitrary date out of the range of ages the oldest known fossil on the monotreme side of the split could have. The other ages were only used as minimal ages; maximal ages were not used, thus providing no protection against too old divergence date estimates. Of course maximal ages cannot be derived from the fossil record as easily as minimal ages, but the fossil record from some epochs is good enough to let us interpret absence of evidence as evidence of absence with more than 50 % confidence.

    There are only three outgroups, increasing the risk that the lengths of these branches could have detrimental effects on the branch lengths in the ingroup…

    No, it is not an earth-shattering paper, despite having been published as an “Article” (rather than “Letter”) in Nature.

    The KT extinction was devastating for just those animals we are attuned to recognize most readily, but may not have been quite so troubling to others; the bacteria didn’t even notice, insects would have gone on buzzing (except those with dependent parasitic relationships), plants may have seen one minor class of leaf-nibblers disappear, the rat-like and lizard-like creatures of the undergrowth might well have found their environments relatively unchanged. That transient period of intense trauma might not have been as disturbing to the overall life of the planet as we big apes might like to imagine.

    Not so. Studies of bite, mining, etc. marks on leaves around the Cretaceous-Paleogene boundary have shown that herbivorous insects suffered massively. Among mammals, one of the three groups of multituberculates disappeared, as did several clades of relatives of marsupials and placentals, and all but apparently one of the dryolestoid fauna of South America (Dryolestoidea being Yet Another Mammal Clade). The lizards were (as expected) hit less severely, but still suffered (for example a herbivorous clade died out). The turtles were decimated enough that a close relative of the alligators that was apparently a specialized turtle-eater died out. And so on.

    And don’t get me started on the Strangelove Ocean…

  2. #2 David Marjanovi?
    March 29, 2007

    I should also have mentioned that, among the plants, the low-latitude conifers (Cheirolepidiaceae) vanished completely. There’s also a mass extinction in angiosperm pollen types.

  3. #3 David Marjanovi?
    March 29, 2007

    But it seems like these days we keep finding Cretaceous ancestors of many extant orders.

    Such as?

    In the meantime, I downloaded the first part of the supplementary information. Table 3 of lists the calibration points. It contains (at least) the following mistakes:

    - Monotrematum is probably not a crown-group monotreme. Thus, Monotremata may be younger than 63.6 Ma.
    - Tribosphenomys is far from being a crown-group rodent! Rodentia (unlike Rodentiamorpha) is younger than 57.25 Ma (…and I wonder where they got such a precise date…). Note that the reference Bininda-Emonds et al. cite makes this quite clear (it coins the name Rodentiamorpha).
    - Not that I knew anything, but I’d be quite surprised if Eodendrogale were a crown-group scandentian. Thus, the crown-group of Scandentia may well be younger than 44.5 Ma.
    - Same for Dendrotherium being a crown-group dermopteran.
    - The genus Tarsius is 44.5 Ma old? Difficult to believe. I’d say a name change is in order.
    - Diacodexis is a paraphyletic series of species around the base of Cetartiodactyla (which means that at least some of those species should get new genus names). Putting “Diacodexis sp.” into Suiformes does therefore not look defensible to me, but I’ll try to find the reference.
    - Obviously, Pakicetus is not a crown-group whale. The crown-group of Cetacea (Autoceta) is much younger.
    - Eomanis is far from a crown-group pangolin…
    - I’d be very surprised if Ageinia turned out to be a crown-group bat. Thus, chances are high that the crown-group of Chiroptera is younger than 52.2 Ma.
    - I’d be surprised if Riostegotherium turned out to be a crown-group dasypodid. But that’s outside my area of knowledge.
    - The crown-group of Tubulidentata consists only of Orycteropus. Obviously Myorycteropus can’t be part of that…
    - Nortedelphys was indeed described as a didelphimorphian, but I don’t buy it. It’s “the tooth, the whole tooth, and nothing but the tooth” (as usual in the Mesozoic). In any case, considering it a crown didelphimorphian really stretches it.
    - Paucituberculata… all extant paucituberculates are caenolestids, and the fossil is most likely not one of those…
    - I bet the fossil notoryctid is outside the crown-group.

    Much sloppier work than I imagined.

  4. #4 David Marjanovi?
    March 29, 2007

    and my point is that seeing a mammalian radiation prior to the K/T boundary should not be a surprising result –

    But a placental and a marsupial radiation in the Cretaceous are surprising results, and therefore I’m not surprised to find so much miscalibration in the paper. Both groups are currently completely absent from the Cretaceous (except for Nortedelphys and a few more problematic crumbs), no matter how common the remains of other mammals are, and no matter how small those are.

  5. #5 David Marjanovi?
    March 29, 2007

    Sorry for a dumb newbie question – but why did the dinosaurs die out completely (including marine and flying dinosaurs), and not reptiles, birds, mammals, etc?

    Those others died out almost completely. :-) The survivors seem to have had the common advantage that they didn’t directly depend on green plant parts or on herbivores.

    BTW, neither the marine nor the flying ones (except the birds) were dinosaurs!

    (Sorry I didn’t mention the birds earlier. Mesozoic birds were a wonderfully diverse bunch. Cenozoic birds — from ostrich to penguin to vulture and back — are all the same by comparison.)

    as far from Yucatan as Australia.

    And New Zealand, and Japan, and and and… :-)

  6. #6 David Marjanovi?
    March 29, 2007

    David – is there another use of the term “crown” in the picture?

    No. For example, you’ll see they put divergence between Tachyglossidae and Ornithorhynchidae ( = the origin of crown-group Monotremata) very shortly after the K-Pg boundary; clearly the Paleocene Monotrematum was considered an ornithorhynchid (or a tachyglossid, but there’s no way to compare two isolated teeth to an echidna, so they clearly didn’t do that).

    A complete account and explanation of what survived and what didn’t is not possible yet, but David will doubtless relate what can be related.

    If possible I’ll avoid it. I’d get stuck in details and keep writing for hours. :-)

    David, others: Why did some lineages survive, and others didn’t? In general, I think you will find that bigger isn’t better. [...]

    Correct. Except maybe for some terrestrial crocs, nothing terrestrial heavier than 25 kg or something survived, probably for the reasons you mention.

  7. #7 David Marjanovi?
    March 29, 2007

    Talk about serendipity in choosing model animals, other than primates, for biomedical research when none of this was known.

    Quite so!

    (The general shape of the tree by Bininda-Emonds et al. is quite robust; since 2001 all molecular analyses have found that shaped, except when they were so plagued by long-branch attraction that the authors themselves pointed that out.)

    However the lack of fossil evidence for obviously diagnosable crown group members of extant orders does not disqualify the molecular conclusions out of hand. It doesn’t mean the lineages didn’t exist either as ghost lineages

    That’s of course correct. But we have pretty large mammal faunas from several places in the world in the Late Cretaceous; the fact that they all lack placentals and marsupials makes it quite improbable that those groups were actually present in the diversity predicted by Bininda-Emonds et al..

    Of course, several Cretaceous mammals have been suggested at one point or another to be placentals or marsupials. For all of those (except the recently described Nortedelphys and friends) alternative suggestions are at least as well supported, however.

    or as morphologically nonderived lineages that appear to be members of stem clades.

    Here, however, you (and plenty of molecular biologists, unfortunately) are saying that all fossilizable innovations of the crown-groups evolved at least twice within the crown-group in question. This is mighty unparsimonious.

    What about birds?

    The fossil record of birds is too bad to tell exactly when all those groups died out, but lots of bird clades were present in the Late Cretaceous and have not been found in younger strata.

    However, crown-group birds were present in the Cretaceous. For example, Vegavis is a crown-group anseriform, more closely related to the ducks + geese + swans than to the screamers and even the magpie goose. That said, Anseriformes is close to the base of the bird crown-group (Neornithes).

  8. #8 David Marjanovi?
    March 30, 2007

    Wow, that is a huge mistake. Are we sure they weren’t using Pakicetus to give a minimum age for the divergence of the Cetecea *stem* group? That would make a lot more sense.

    Yes, but look at the tree: they put the divergence between baleen whales and toothed whales into what seems to be the Eocene. Or have a look at their calibration point for Monotrematum, the only known Paleocene monotreme — the divergence between Tachyglossidae and Ornithorhynchidae is right after the K-Pg boundary, fitting the age of Monotrematum, so Bininda-Emonds et al. seem to consider M. an ornithorhynchid.

    This study would be more impressive if they had compared the crown group frequency to some kind of null model.

    I agree.

    Note how little time passed between reception and acceptance (just over 3 months) and between acceptance and publication (just under 2). That’s very fast.

    The relative lack of placentals and marsupials in known mammal faunas from the late cretaceous doesn’t really seem tome to provide a strong case that such groups did not exist. Assuming Eomaia scansoria is a proper Eutherian

    Indeed it is. It is close to the base of Eutheria, however, while Placentalia is the crown-group of Eutheria. It can be used to calibrate the split between Placentalia and Marsupialia (or rather Eutheria and Metatheria), which Bininda-Emonds et al. did; it has no bearing on the time of origin of Placentalia.

    If there are comparatively few fossil examples of eutherians in the mesozoic record

    This, however, is not the case. Non-placental eutherians are pretty common in the Late Cretaceous of central Asia (from Uzbekistan to Mongolia), western North America, and Europe (though very few Late Cretaceous mammals are known from Europe in total, so this may not count). Possible placentals (cimolestans, taeniodonts) are rare, and certain placentals are absent so far.

    The same holds for non-marsupial metatherians and marsupials.

    Proto-placentalians may have been minor members of the fauna, as Mr. Marjanovic indicates, but they still could have been divided into several lineages that gave rise to different modern placental orders after the age of the dinosaurs.

    This is of course correct; neither the fossil record nor our current knowledge of it are complete. But have another look at the tree: I count not two, not five, not ten, but forty-two lineages of Cretaceous placentals and marsupials, and those are just those that happen to have left extant descendants. That so many lineages were present in the Cretaceous but are so far absent from the fossil record is not probable. (I can’t do the math off the top of my head, but it can be done — a paper has recently been submitted which does something like this for amphibians.)

    in seeming ignorance of the tectonic timetable for isolation of South America from Australia.

    Granted, the entire terrestrial fossil record from the Late Cretaceous of Australia is currently limited to one dinosaur bone. Anything could have been present or absent. And you are right that the didelphimorphians and paucituberculates are South American marsupials.

    However, you have overlooked four important things.

    One is the fossil record of South America. Plenty of mammals are known from its Late Cretaceous layers. What mammals? Dryolestoids all over the place, plus a few gondwanatheres. That’s it. No metatherians, no eutherians except maybe one tooth.

    Another is the fossil record of Madagascar. South America and Madagascar had almost the same dinosaurs in the Late Cretaceous because they were connected for much or maybe all of that time via Antarctica, India, and the Seychelles. There are fossil mammals known from the Late Cretaceous of Madagascar: a gondwanathere, an unpublished weirdo known from a complete skeleton or something, and one tooth that was initially interpreted as a metatherian but has been reinterpreted as a probably non-placental eutherian.

    This brings us to India. One gondwanathere; a few teeth and ankle bones of what seem to be non-placental eutherians.

    Now for the fourth. Metatheria did not originate in Gondwana but in Laurasia. There’s a plethora of Mesozoic metatherians in Asia and North America, from the very base of their tree (Sinodelphys) to the end-Cretaceous “pediomyids” (shrew-like) and stagodontids (Tassie-devil-like). Indeed, superficially opossum-like metatherians survived the K-Pg boundary, temporarily even spread into Africa, and died out in the northern continents pretty late in the Miocene. In South America and Australia, Marsupialia (and the slightly larger group Notometatheria as a whole) appears in the Paleocene. Clearly, the ancestors of Notometatheria came from North to South America sometime around the K-Pg boundary (perhaps shortly before, as for example the hadrosaurid dinosaurs did), then had their initial radiation there and entered Australia via Antarctica in the early or middle Paleocene.

    Note that the authors of this paper do not “show” that placentals were diversifying in the mid Cretaceous, they have *extrapolated* this conclusion from their data, which is not quite the same thing.

    This extrapolation would be pretty robust if the calibration points were reliable and if the effect of maximal ages had been explored…

    I’ll try to get your paper, thanks for alerting me. :-)

  9. #9 David Marjanovi?
    March 30, 2007

    I should have mentioned that the only certain fossil ornithorhynchids are Miocene and younger. There are no known fossil tachyglossids (apparently the group evolved in the underexplored New Guinea). There are no known fossil monotremes from the Eo- or Oligocene (or for that matter the Late Cretaceous… there are several from the Early Cretaceous, however). Thus, Bininda-Emonds et al. have obviously used Monotrematum to calibrate the divergence between Ornithorhynchidae and Tachyglossidae.

  10. #10 David Marjanovi?
    March 30, 2007

    Oops, there are known fossil tachyglossids, but only from the Pleistocene and (I forgot) maybe the Pliocene, which doesn’t change my point.

  11. #11 David Marjanovi?
    March 30, 2007

    Oops… I didn’t quite answer the question. Complete separation of Australia and Antarctica, eastern Antarctica icing over: apparently 45 Ma ago (mid-late Eocene). Separation of South America and Antarctica, western Antarctica icing over: some 37 Ma ago (late Eocene). Earliest marsupials in Australia: over 55 Ma ago (late Paleocene). Earliest marsupials in South America: up to 65 Ma ago (early Paleocene).

    The Hunter & Janis paper is here; even if you don’t have full access, you should be able to read the abstract. (I have read the paper a couple of months ago. It’s very good. Of course, it inevitably suffers from the continuing lack of a morphological analysis of placental phylogeny of reasonable size.)

  12. #12 David Marjanovi?
    March 30, 2007

    This HAS been done — see Foote et al. Science 1999.

    Yes, but it should probably be updated. (I don’t think the conclusion would change, though.)

    I agree with post 46, Repenomamus giganticus from the middle Early Cretaceous being the exception that proves the rule.

    Is it possible that the meteor strike killed the large plants.

    This seems to be what happened, but keep in mind that trees are “large plants”, and that you can’t grow when you’ve starved to death… or what do you mean? ~:-|

    As far as I am aware, zalambelestids and zhelestids are probably the closest outgroups to placentalia we currently know of

    Things like Cimolesta and Leptictida are probably much closer, and present in the last stage of the Cretaceous but AFAIK not earlier. (They are seemingly still not inside the crown-group, but there’s no good phylogenetic analysis of that part of the tree…)

    If this is so, then one could take the view that the _initial_ inerordinal divergences (i.e. between afrotheria and xenarthra) among placentalians could have occurred anytime between 65 million years ago and 90 million years ago, and maybe even a little further back.

    Fine — but 42 such divergences is clearly too much.

    Also, aside from this, I suspect the role of giant birds as the primary predator in many systems during the Paleocene results in some of the patterns we see.

    The latest word on the gastornithids — Diatryma is a junior synonym of Gastornis — is that they were herbivores, eating e. g. palm hearts.

    it appears from the comments that the relationships shown in the tree are fairly accurate, just the dates in question.

    Yep. Part of the reason is that most of the tree is simply not new — that’s why it’s called a supertree: it’s a combination of trees that have already been published elsewhere.

    I find it very curious that the bovines and cervines are closer to the cetaceans than to the equids, and that all of those are closer to the carnivora and chiroptera that to the rodentia!

    That’s normal. Every molecular analysis that doesn’t suffer from obvious problems has found these relationships (using different combinations of different genes and different methods). Morphological analyses have since confirmed the sister-group relationship of hippos and whales (leaving the position of the ruminants as the only difference between morphology and molecules in the cetartiodactyl tree) and are so far inconclusive on the rest, though AFAIK monophyly of all herbivorous placentals has never been suggested.

    What do elephants and whales have in common? Did you confuse the whales with the sea cows?

    On the phylogenetic position of the bats, I’ll try to dig up the link to the Pegasoferae paper (which is free-access). Pegasoferae is what the clade composed of Perissodactyla, Chiroptera and Carnivora (and by implication Pholidota) has been named. It is the sister-group of Cetartiodactyla. :-)