Field of Science

The Font of the Placentals

The large-scale incorporation of molecular data into phylogenetics over the last few decades has caused a revolution in our understanding of life's evolution. Taxa whose interrelationships were previously regarded as intractable have been opened up to study, and many of our previous views on relationships have been forced to shift. Because conflict always makes for a good story, certain cases of the latter have become causes celebres, receiving extensive attention in both the technical and popular literature. One of these subjects of particular interest, not surprisingly, involves the relationships of the living orders of mammals.

Reconstruction of Arctostylops steini by Brian Regal, from Janis et al. (1998). The arctostylopids are a Palaeocene to Eocene group of mammals of uncertain affinities but probably belonging somewhere in the Boreoeutheria.

A lot of this attention has focused around the revelation of the Afrotheria, a grouping of animals (tenrecs, elephant shrews, hyraxes, aardvarks, elephants and manatees) with likely African origins that was completely unsuspected by studies based on morphological data only but which molecular studies have identified with ever-increasing levels of support. Recent molecular studies of placental phylogeny have agreed on three basal divisions within the placental mammals: the Afrotheria, the Xenarthra (armadillos, anteaters and sloths, a grouping that was recognised even before the advent of molecular data), and the remaining placentals in the largest of the three, the Boreoeutheria.

To the best of my knowledge, the Boreoeutheria is a clade that has also so far been supported by molecular data only with no morphological features yet recognised as defining the group. Nevertheless, its support can be considered as well established. The name Boreoeutheria refers to the clade's likely northern origins in contrast to the more southern distribution of the other two. Within the Boreoeutheria, molecular studies indicate a basal divide between the Euarchontaglires on one side and the Laurasiatheria on the other. The Euarchontaglires include the primates and rodents (as well as a handful of smaller orders). The Laurasiatheria include the Eulipotyphla, a group of insectivorous mammals including shrews, moles and hedgehogs, as sister to a clade containing bats, carnivorans, perissodactyls and artiodactyls.

Molecular phylogeny of mammals, from Springer et al. (2004) (note that not all branches shown in this tree are supported by all studies).

This all has interesting ramifications for the early evolution of placentals. There is an extensive fossil record of mammals from the Palaeocene, the epoch of time immediately following the end of the Cretaceous. However, most of these mammals do not belong to the orders alive today and their exact relationships to living mammals remain open to debate. The molecule-induced shake-up of pacental relationships just increased this uncertainty: for instance, the interpretation of a given group of fossil mammals as close to the common ancestry of perissodactyls and elephants rather goes out the window when perissodactyls and elephants are no longer thought to be closely related. And detailed studies that may resolve these issues remain few and far between. One of the most notable analyses in recent years has been that by Halliday et al. (2017) which covered most of the well-preserved placentals and their close relatives from the Cretaceous and Palaeocene periods. However, it is difficult to say just what to make of their results. The unconstrained analysis of their data presents results that remain deeply inconsistent with the molecular tree. Conversely, constraining the analysis to more closely match the molecular data provides results that are intriguing but difficult to accept at face value; I suspect they may be artefacts of the algorithm forcing taxa into the least unacceptable position for inadequate data. Suggesting that pangolins are the last specialised survivors of a broad clade of condylarths, pantodonts, notoungulates and creodonts is... I suppose not a priori impossible, but definitely a big call. A later analysis based on an expanded version of the same data set by Halliday et al. (2019) irons out some of the kinks but still fails to resolve the base of the Boreoeutheria beyond a massive polytomy of 25 branches (an icosipentatomy?). The Euarchontaglires are recovered as a clade but not the Laurasiatheria or any of its molecular subgroups above the ordinal level. And while some of the newer analysis' placements may seem like an improvement (notoungulates are placed as the sister to litopterns instead of hanging out with pangolins), others may still raise an eyebrow (mesonychids are associated with carnivorans but viverravids and miacids are not).

As always, the best answer to this conundrum is likely to involve more research. While researching this post, I did come across comments from people suggesting issues with the Halliday et al. data. Frankly, for a data set of this size (involving 248 taxa and 748 characters in the 2019 paper), it would be incredible were it otherwise. I know from my own experience that as you add more characters and taxa to a phylogenetic analyses, the challenge of keeping everything in line rises exponentially, and the data sets I've dealt with have been nowhere near the size of this one. Nevertheless, it's a start. And we can but hope that even those who find fault with it ultimately take it as inspiration to themselves do better.


Halliday, T. J. D., M. dos Reis, A. U. Tamuri, H. Ferguson-Gow, Z. Yang & A. Goswami. 2019. Rapid morphological evolution in placental mammals post-dates the origin of the crown group. Proceedings of the Royal Society of London Series B—Biological Sciences 286: 20182418.

Halliday, T. J. D., P. Upchurch & A. Goswami. 2017. Resolving the relationships of Paleocene placental mammals. Biological Reviews 92 (1): 521–550.


  1. Have there ever been studies that test this method of forcing molecular constraints on morphological data? For instance, leaving out one order of mammals at a time from the constraint backbone tree and seeing if it actually emerges in the position shown by genetics. If these "pretend extinct" orders don't emerge in the correct place, then I think we'd have some big problems with the phylogeny of extinct taxa.

    1. Springer et al. (2007) tried this, but unfortunately their results weren't promising. It does seem to work pretty well in some cases, especially assuming high completeness of fossil specimens, as found by Pattinson et al. (2015) regarding primates.

      More recently, Beck and Baillie (2018) used an alternative method in which morphological traits of "hypothetical ancestors" were inferred based on molecular topologies of placentals, and by including these ancestors in analyses were able to recover a topology similar to molecular topologies with only morphological data. Asher et al. (2019) applied this method to rodents, also with overall favorable results.

    2. I was just coming here to mention Beck & Baillie, but I see I was beaten to it. Obviously, this isn't my field, but my impression is that things are more stable in parts of the tree than others. Euarchontaglires seems to work fairly well, and interestingly enough, Afrotheria seems to be playing ball more as more basal taxa get added in. Laurasiatheria, however, remains pretty intractable, probably because modern representatives are either super conservative (Lipotyphla), super derived (Artiodactyla, Perissodactyla, Pholidota), or somehow both super conservative and super derived at the same time (Chiroptera, Carnivora).

    3. Springer et al. (2007) is exactly what I was thinking and it failed miserably- "only 5 of 18 orders (rodents, lagomorphs, scandentians, sirenians, proboscideans) consistently remained in the molecularly defined superordinal group." But it's also a tiny matrix (196 characters). I'd be interested to see how O'Leary et al.'s 4541 character matrix featured in Beck and Ballie's paper did, since it's good enough to recover Xenarthra, Cetartiodactyla, not Volitantia, etc.. But I would absolutely not trust placing extinct taxa in a molecular backbone in Asher's matrix.

      I'm not sure about Beck and Baillie's method. I think if you're using taxa with so much separation to infer ancestral traits, you're going to be wrong quite often due to convergence in living taxa. Picture the hypothetical common ancestors along the maniraptoriform backbone if you were just using Late Cretaceous Dromiceiomimus, Shuvuuia, Segnosaurus, Elmisaurus, Troodon and an avian as your OTUs. Premaxilla ancestrally toothless, maniraptorans ancestrally opisthopubic, metatarsus ancestrally arctometatarsalian, etc.. So something like Falcarius or Caudipteryx could be shoved out of the group, and I think you see that in B&B's tree with that huge crowd of archaic ungulates and Rodhocetus sitting outside crown Ungulata. All the extinct Glires being stem lagomorphs raises a similar flag.

    4. I also wonder if Beck & Baillie's method could be accused of being circular: if we add hypothetical taxa with an ideal character distribution for recovering our favoured tree, then we recover our favoured tree. As you say, Mickey, the common ancestor of two living taxa will probably not be a perfect medium of the two.

      And you're right about the Asher matrix. Asher et al. (2003) was the analysis that suggested a number of 'condylarths' were afrotherians rather than euungulates, whereas more recent, more extensive data sets have shifted them towards a more traditional position.

  2. The big O'Leary et al. (2013) study did identify some morphological synapomorphies for Boreoeutheria (in their supplement), but perhaps unsurprisingly, many of those features have undergone reversals within the group.


Markup Key:
- <b>bold</b> = bold
- <i>italic</i> = italic
- <a href="">FoS</a> = FoS