Tuesday, June 10, 2014

The best known fossil pinniped, part 4:Phylogenetic relationships and evolution of Allodesmus, and future directions

Finally, we're at the last post in this series - one that I've been procrastinating on, as I knew it would be a total bear to write. But, now I'm done! Hooray. Next up after this - a recent announcement of fossil marine vertebrate discoveries in the East Bay (e.g. east San Francisco Bay Area) and some photos from a recent field excavation, or as I like to call it, going fishing in the Oligocene.

The phylogenetic position of Allodesmus

The relationships of Allodesmus have been a serious bone of contention (pun intended…sorry) in pinniped paleontology for several decades, and is a bit of a polarizing issue within the field. I have my own biases which no doubt will shine through below, but will try to show both sides of the argument. Two major camps exist within pinniped paleontology: those in favor of pinniped diphyly and otarioid monophyly (Otariidae + Odobenidae + Desmatophocidae + enaliarctines), and those who embrace pinniped monophyly and phocoid monophyly (Phocoidea is a Desmatophocidae + Phocidae clade).

A handy hand drawn phylogenetic tree with illustrations of skulls. From Barnes et al. (1985). Although outdated, no subsequent illustration has ever been put together of this caliber for pinniped evolution, and I've always wanted to redo this given our substantially expanded fossil pinniped record.
Mitchell (1966) arguably represents the first “modern” treatment of Allodesmus, and while he did not explicitly propose any phylogenetic hypothesis for pinnipeds, he  recognized that Allodesmus was quite derived and probably not directly ancestral to living sea lions. Prior workers, in the absence of other fossils, implicated the fragmentary remains of Allodesmus in the ancestry of modern pinnipeds. We now know that desmatophocids were a relatively highly derived evolutionary experiment that lasted for about 10 Ma – and ultimately went totally extinct – and despite predating true seals (by 5 Ma), walruses (by 2-3 Ma), and fur seals and sea lions (by 10 Ma or so) – desmatophocids evolved independently from early pinnipeds and did not give rise to any of these modern groups. Given the sheer abundance of Allodesmus in the mid Miocene fossil record, it is ultimately unsurprising that it was one of the earliest discovered fossil pinnipeds; unfortunately, this also remained one of the only well known fossil pinnipeds for quite some time, and it was not until the 1970’s that paleontologists could make heads or tails of it.
Allodesmus was classified as an otariid relatively early on by such researchers as Remington Kellogg, Theodore Downs, and Victor Scheffer. Mitchell (1966) suggested some relationship with walruses and sea lions, and erected the subfamily Desmatophocinae to include it and Desmatophoca (and curiously, the walrus Dusignathus) within the Otariidae (=Otarioidea of other workers). It must be stated that during the 1960’s, 1970’s, and 1980’s, most pinniped researchers were working under the implicit assumption that pinnipeds were diphyletic – with walruses and sea lions sharing a common bear-like ancestor, and true seals evolving separately from a mustelid ancestor (more on this later).

The phylogenetic hypothesis of Barnes (1972). 
Barnes (1972) recognized that desmatophocids diversified prior to the appearance of otariids, and even recognized that otariids were more primitive than desmatophocids in many regards, despite diversifying long before the appearance of the earliest fossil otariids. Barnes (1972) was appropriately skeptical that otariid ancestry had anything to do with desmatophocids, and suggested that otariids were secondarily derived from a primitive ancestral stock, and similarly indicated that desmatophocids could probably not be implicated in walrus origins either. Barnes’ (1972) proposed phylogeny can be seen above. Mitchell (1975) published a phylogenetic hypothesis which did not differ appreciably from Barnes (1972).

The phylogenetic hypothesis of Mitchell (1975). 
A later study by Repenning and Tedford (1977) showed a more or less similar phylogeny, with desmatophocids and odobenids sharing a more recent common ancestor than otariids. Repenning and Tedford (1977) further agreed that no modern pinnipeds could have been derived from the desmatophocids.

The phylogenetic hypothesis of Repenning and Tedford (1977). 
New research published by Andre Wyss in 1987 indicated the possibility that pinnipeds were not only monophyletic (spurred by early studies of molecular phylogeny), but that walruses and true seals (together with desmatophocids) might form a monophyletic group (the clade Phocomorpha). As a response to this, Barnes, in his 1989 description of the enaliarctine Pteronarctos goedertae, published a new phylogeny for fossil and modern pinnipeds (see below). Curiously, one of the earliest comments regarding Wyss’ hypothesis of pinniped monophyly was that his “conclusion is contradictory to that reached by many other researchers who have examined many of the same characters as Wyss”. Although a fair comment here, this point has been reiterated time and again by proponents of pinniped diphyly, in print and in person to the point where it is reminiscent of an appeal to authority. But I digress. Barnes (1989) shows a hand-drawn cladogram depicting otarioid relationships, and pointed out a couple of minor holes in Wyss’ arguments, including the observation that phocids, desmatophocids, and the enaliarctine Pinnarctidion all share a “mortised” (e.g. vertically thickened/expanded) zygomatic arch (or cheekbone for the uninitiated). Barnes pointed out that some odobeninae also possess this in addition to sirenians and desmostylians, and that the convergent nature of this feature means it should be avoided in phylogenetics. Such arguments are commonplace in vertebrate paleontology, and the idea that certain types of characters ought to be avoided because they are convergent is pervasive. However, the moment any character evidence is ignored – for any reasoning – is the moment that subjectivity is introduced into the analysis. Conversations with Jonathan Geisler back in January 2011 opened my eyes to this, and as a result, my own philosophy is to include as much character evidence as possible.

The phylogenetic hypothesis of Barnes (1989). 
Regardless, Barnes (1989) did not present a computer-assisted analysis, but rather listed a series of characters supporting otarioid monophyly, and other clades within the Otarioidea. In a non-computer assisted analysis, a hand-drawn cladogram is assembled based on the listing of characters supporting “premeditated” clades – no testing is involved, and it is completely subjective. The 1980’s saw the proliferation of computer-aided cladistic analysis (referred to hereafter as just cladistic analysis). For the uninitiated, cladistic analysis emphasizes the grouping of organisms based on shared features rather than differences. This sort of analysis is conducted by producing a list of characters (e.g. characteristics or features) with different character states (e.g. conditions). An example of a classic (and already mentioned) character for fossil pinnipeds would be:

Character 1: zygomatic arch. 0-splintlike, zygomatic process tapers. 1-mortised, distal zygomatic process is dorsoventrally expanded.

This character would be coded as such:

Grizzly bear (Ursus): 0
Enaliarctos (earliest fossil pinniped): 0
Northern Fur Seal (Callorhinus): 0
California Sea Lion (Zalophus): 0
Walrus (Odobenus): 0
Desmatophoca: 1
Allodesmus: 1
Harbor Seal (Phoca): 1
            All species coded as ‘1’ would be grouped by the computer into a clade. In this case, the supported clade would be the clade Phocoidea. It is possible to have more than two character states; only derived character states (e.g. anything other than zero) will unite species together. When an analysis includes hundreds of characters like this, it is likely that some will be mutually exclusive: for example, in the context of this discussion, some characters might support otarioid monophyly and others might support phocoid or phocomorph monophyly. What happens when two characters support mutually exclusive groups of species? The short answer is that, as you might suspect, they cancel each other out. At larger scales, if an analysis includes dozens of characters that support mutually exclusive clades – they will also cancel each other out if they are of roughly equal proportions. This can be solved by one of two ways: ignoring characters that support a clade you don’t like, or searching for additional characters that might support one hypothesis over another. Admittedly, the first option sounds dangerously unscientfic (…because it is), but plenty of paleontologists are guilty of this practice, unfortunately.

An excellent summary figure contrasting the "newly" proposed phylogeny under the "Phocomorpha hypothesis" of Berta and Wyss (1994) with the "classical" view of pinniped diphyly advocated by most earlier researchers. From Berta and Wyss (1994).
            A response to the phylogenetic hypothesis of Barnes (1989) was a large cladistic analysis of morphological data by Berta and Wyss (1994) that still remains the only comprehensive phylogenetic analysis of pinnipeds (as alluded to below, Morgan Churchill and I presented an even larger analysis at the 2010 SVP meeting, but it needs a lot more work and was a good “first stab”; comprehensive phylogenetic analysis of pinnipeds remain a viable future research project). This study used 143 characters including skeletodental, soft tissue, and even behavioral characters. Berta and Wyss (1994) found strong support for a monophyletic Pinnipedia, monophyly of the Phocomorpha (walruses + desmatophocids + true seals), and monophyly of the Phocoidea (Desmatophocidae + true seals + Pinnarctidion). Berta and Wyss (1994) identified that the sheer majority of purported synapomorphies for the Otarioidea also occurred in terrestrial arctoids (indicating that they are primitively present in “otarioids” and are thus not phylogenetically informative), and that many other features are also present in true seals (which, of course, were not considered by Barnes because of the a priori assumption that pinnipeds are diphyletic).

The phylogenetic hypothesis of Berta and Wyss (1994); black bars denote the number of synapomorphies for each node (I think). 
            A biting response to this analysis was published less than a year later by Barnes and Hirota (1995) who criticized the definition of a few features (some of which were later acknowledged by Deméré and Berta, 2002, to be valid criticisms). Barnes and Hirota (1995) made some comments on Berta and Wyss characters which in my mind don’t make much logical sense. For example, they criticize the use of a character based on the presence of an enlarged digastric insertion on the mandible as it is “related to the musculature and functional morphology of the dentary and not of phylogenetic significance” (Barnes and Hirota, 1995:356). So what? I fail to see the distinction here: all morphological features are influenced by function to some degree, and either characteristics have a known function, or the function is unknown and we simply do not know enough to make an informed statement about the functional significance of a certain feature. The second puzzling statement is a criticism of the use of soft tissue features by Berta and Wyss (1994), as “soft anatomical structures that are not reflected by bone morphology cannot elucidate the relationships of fossil taxa”. While it is true that such soft tissue features are not preserved in desmatophocids, they are present in the walrus, which is sister to the desmatophocid + true seal clade (Phocoidea) in the analysis of Berta and Wyss (1994). Here’s the problem with this statement: when it comes to cladistic analysis, soft and hard tissue characters hold no inherent value over one another: they are all equally important according to the computer. To the computer, a soft tissue character codable only for modern pinnipeds is no different from, say, mandibular characteristics that can only be coded for fossil and modern species with preserved mandibles: no distinction is made. Thus, the argument that a certain character is irrelevant because it can only be coded for a small number of species is erroneous. To further illustrate this, there are some skeletal characteristics that could potentially be coded for fossil pinnipeds – if found beautifully articulated – that can generally only be coded in modern specimens. Subsequent studies have focused on smaller groups of pinnipeds, and not a single morphological study since has really tested any of the larger clades within pinnipeds – and that includes the phylogenetic position of Allodesmus as a true seal relative.
            So, where does that leave us today? It’s now 20 years since Berta and Wyss (1994) was published, establishing a new paradigm for pinniped phylogeny – and we are still without a followup to it. It is an excellent paper, and I strongly recommend that any student of pinniped evolution read it. However, there’s a serious catch. Virtually all published studies of molecular data have unilaterally supported an otariid-odobenid clade (Otaroidea – walruses and sea lions), and also supporting pinniped monophyly. So, molecular phylogeny suggests a bit of a compromise between the two extremes. But, molecular analyses – like soft tissue data – are inapplicable to the question of where Allodesmus fits in. The postcranial skeleton of Allodesmus is very similar to odobenids and otariids – while many features of the skull are more similar to true seals. Personally, I’ll remark upon the fact that there are far fewer skeletal features linking walruses with Phocoidea than there are supporting the Phocoidea (desmatophocids + true seals), and I could be convinced of otarioid monophyly if someone were to carefully identify a suite of synapomorphies. In this context, my own suspicion is that molecular work is probably correct – but as of yet I cannot point to much morphological evidence in support of it – and that Allodesmus most likely, given the volume of shared features identified by Berta and Wyss (994), does share common ancestry with the true seals. Much, much more work is necessary to test all of these hypotheses, and I am not particularly beholden to any idea, other than it is abundantly clear that pinniped diphyly is an outdated concept not supported by modern science and lacking molecular and anatomical support. Another post on pinniped diphyly would be instructive.

Evolution of Allodesmus

Regardless of larger phylogenetic relationships, a few comments can be said about the evolution of this fossil pinniped are worth making. Allodesmus is the sister genus of Desmatophoca – and if a genuinely separate genus, Atopotarus is closer yet to Allodesmus. Desmatophoca is fairly old – the type specimen of Desmatophoca brachycephala is perhaps as old as 21 Ma, demonstrating that desmatophocids diverged fairly early from enaliarctines. Desmatophocids in general are noteworthy for being the earliest large bodied pinnipeds to evolve; both species of Desmatophoca are large, and most species of Allodesmus are even larger. Numerous changes occurred between the evolution of Allodesmus from Desmatophoca-like morphology (fossils from the Astoria Formation purportedly show that Desmatophoca oregonensis and an “Allodesmine” co-occurred, thus demonstrating that the two do not share an ancestor-descendant relationship – but this undescribed Allodesmus/Allodesmine has yet to be described; Barnes and Hirota, 1995). Changes in the skull and mandible include elongation of the rostrum, enlargement of the orbits, development of a prenarial shelf, further elaboration of the “mortised” zygomatic arch, fusion of postcanine tooth roots, simplification of tooth crowns, elongation of the mandible, and enlargement of the posteroventral flange for the digastric insertion of the mandible.
            Curiously, Allodesmus appears to have gone extinct by the early Messinian or late Tortonian (~9 Ma), based on the age of the youngest known specimens from the Montesano Formation of Washington and the Santa Margarita Sandstone near Santa Cruz. Post-middle Miocene specimens of Allodesmus are also relatively rare: the post-middle Miocene record is effectively restricted to the partial skeleton from the Montesano Fm. of Washington, about a dozen teeth from the Santa Margarita Sandstone near Santa Cruz, and a few teeth from the Monterey Formation of Orange County I saw at the Cooper Center last Fall. In contrast, the late Miocene record of larger pinnipeds is dominated by “imagotariine” and dusignathine walruses like Imagotaria, Gomphotaria, Dusignathus, Pontolis (and two other unnamed dusignathine genera) and the early odobenine Aivukus. Other pinnipeds include the early otariids Pithanotaria starri and Thalassoleon. Interestingly, early and Middle Miocene walruses (e.g. Proneotherium, Neotherium, Pseudotaria) that co-occur with desmatophocids are mostly relatively small bodied, similar in size to large fur seals and female sea lions (Pelagiarctos is a notable exception, although it wasn’t very large). However, in the late Miocene, walruses both increase in body size (the “smaller” walruses like Imagotaria were only the size of large California sea lions, while monsters like Pontolis were five or six meters in length – southern elephant seal sized) but also became more diverse.

Future Directions

Clearly, the issue of the phylogenetic position of Allodesmus amongst pinnipeds is not quite a dead and beaten horse, rather one that has been assaulted numerous times but failed to die. As things stand, we know barely more about the relationships of this species than we did in 1990 – which is not to deride either side of the debate. In my opinion, the answer has not yet been adequately addressed, and more character evidence - and a much broader taxonomic sample - are needed. Smaller studies with limited taxonomic sampling are inherently going to be less adequate than broader approaches, and I think the next big step is going to be putting together a comprehensive phylogeny of all modern and fossil pinnipeds (at least, the ones complete enough to be useful, that is). This is something Morgan Churchill and I are planning on in the somewhat distant future; our first stab at this was presented on a poster at the 2010 SVP meeting in Pittsburg, which showed a somewhat similar topology to that of the Berta and Wyss (1994) paper. As we speak, some promising research is being done by Reagan Furbish, a student of Annalisa Berta at San Diego State University. Reagan had a pretty neat poster about her Master’s thesis work on the phylogenetic position of Allodesmus, which actually won an award for best student poster at the Marine Mammal conference here in Dunedin last December. Needless to say I’m really looking forward to seeing what else she digs up.

An undescribed skull and mandible of Allodesmus sp. from the Monterey Formation, on display at the San Diego Natural History Museum. Does this represent a new species of Allodesmus? Or a previously described species? The short answer is, nobody knows (yet). 
In addition to these ongoing issues, there is a whole lot more material of Allodesmus known now. The late Miocene Allodesmus specimen at the Burke Museum (Montesano Formation, Washington) is still undescribed, and there are numerous isolated teeth in collections at UCMP and the Santa Cruz Museum representing similarly young specimens of Allodesmus from the Santa Margarita Sandstone near Santa Cruz (the latter being an easy project I’ll probably take on in search of “low hanging fruit”). Last fall I saw a partial but nonetheless informative skull (with mandibles and an atlas) of Allodesmus kernensis from Sharktooth Hill at Sierra College; this, and other new specimens from SDNHM (including another nice, partially disarticulated skull), certainly warrant further study. If you read Mitchell’s (1966) paper on Allodesmus, yeah, it’s big and thick and well illustrated, but the skull description is actually quite brief. It’s been supplemented by the descriptions in Barnes (1972), but more material always offers to provide new insights. Another undescribed Allodesmus from the Monterey Formation is on display at SDNHM. There is undescribed Sharktooth Hill Bonebed material at LACM that needs publishing, but I can’t say more about it. Barnes and Hirota (1995) also mention an early Miocene “allodesmine” from the Astoria Formation of all places.

Skulls like this (Allodesmus kernensis, Sharktooth Hill Bonebed, SDNHM) are by no means as impressive as some of the previously published specimens - but even squashed, roadkill specimens like this preserve a hell of a lot of morphology and beg - no, deserve careful study. 
Allodesmus has been arguably oversplit, and the taxon is in dire need of revision. Detailed study of new material – whether from Sharktooth Hill or beyond – will help resolve questions of diversity, biogeography, and phylogeny.
What else can be done? The sample from the Sharktooth Hill Bonebed alone is sufficient enough to permit studies of variation of any number of elements (skulls, mandibles, teeth, postcrania), in addition to providing well-preserved specimens for functional analysis (feeding ecology, muscle attachment mapping, further studies of locomotion) and studies of ontogeny and sexual dimorphism, as well as isotope geochemistry (e.g. Oxygen and Carbon isotopes for studying paleoecology). There is literally a host of avenues for future research – just on a single genus of pinniped. The sky’s the limit – and Morgan and I can’t do all of it! I hope this reaches someone who can help out with the California marine mammal fossil record – there is so much more left to do, even on the best known fossil pinniped.


L. G. Barnes. 1972. Miocene Desmatophocinae (Mammalia: Carnivora) from California. University of California Publications in Geological Sciences 89:1-76.

L.G. Barnes. 1989. A new enaliarctine pinniped from the Astoria Formation, Oregon, and a classification of the Otariidae (Mammalia: Carnivora). Contributions in Science, Natural History Museum of Los Angeles County 403:1–26.

L. G. Barnes and K. Hirota. 1995. Miocene pinnipeds of the otariid subfamily Allodesminae in the North Pacific Ocean: Systematics and relationships. The Island Arc 3:329-360.

L.G. Barnes, D. P. Domning, and C. E. Ray. 1985. Status of studies on fossil marine mammals. Marine Mammal Science 1:15–53.

A. Berta and A. R. Wyss. 1994. Pinniped phylogeny. Proceedings of the San Diego Society of Natural History 29:33–56.

T.A. Deméré and A. Berta. 2002. The Miocene pinniped Desmatophoca oregonensis Condon 1906 (Mammalia: Carnivora), from the Astoria Formation of Oregon; pp. 113–147 in R. J. Emry (ed.), Cenozoic Mammals of Land and Sea: Tributes to the Career of Clayton E. Ray. Smithsonian Contributions to Paleobiology 93

E. D. Mitchell. 1966. The Miocene pinniped Allodesmus. University of California Publications in Geological Sciences 61:1-105.

A.R. Wyss. 1987. The walrus auditory region and the monophyly of pinnipeds. American Museum Novitates 2871:1–31.


J. Velez-Juarbe said...

Very nice!! Reagan Furbish, a student of Annalisa, gave a very interesting presentation on Allodesmus at SecAd. Hopefully once a certain collection is more open she'll be able to fully look at the available material :)

Robert Boessenecker said...

Say no more, say no more!

I saw Reagan's poster at the SMM meeting last December here in Dunedin - I'm really looking forward to seeing her master's work in print. It does make me pretty happy to see other people getting interested in fossil pinnipeds.

Ray Pierotti said...

What has happened to your May 12, 2018 post in which you discussed Pinniped evolution? It contained a lot of interesting material, but it is no longer accessible. Attempts to access quotes from it are linked to your Jan 2012 article on Puijila, but the quotes are not in that article.