Breaking the Chain of Early Whale Evolution


A comparison of the third molars from three species of Pakicetus as viewed from the back. (From Cooper et al., 2009)

Crack open just about any recent popular overview of evolution (namely Why Evolution is True, The Greatest Show on Earth, and Evolution: What the Fossils Say and Why it Matters) and somewhere inside you will find a string of skeletal whales. Starting with either Indohyus or Pakicetus, the illustration will feature a graded series of forms that connect modern whales with their terrestrial ancestors. A caveat may be included in the text to say that we cannot be absolutely sure that each included genus gave rise to the next, but the general idea is that the evolution of whales occurred in a gradual, linear fashion through a series of intermediate stages.

Such iconography is not entirely wrong. We know that living whales are the descendants of land-dwelling ancestors that lived about 55 million years ago and the widely-published sequence roughly documents how the ancestors of living whales were adapted to life at sea. But by only being concerned with connecting living whales to their early ancestors, however, we run the risk of suggesting that such illustrations accurately depict the entire evolutionary history of whales. They most certainly do not.

Earlier this year I wrote about a redescription of the early whales Kutchicetus and Andrewsiphius by fossil whale experts Hans Thewissen and Sunil Bajpai. Neither of these forms fit along the neat continuum presented by Coyne, Dawkins, and Prothero in their books. Instead Kutchicetus and Andrewsiphius were part of a radiation of long-snouted, otter-like whales called remingtonocetids which lived alongside forms taken to represent a transitional stage to living whales such as Rodhocetus and Maiacetus. Indeed, if you could travel back about 48 million years ago to the beaches and nearshore environments of what is now northern Pakistan you would encounter a diversity of early whales inhabiting a range of habitats.

A paper just published in the latest issue of the Journal of Vertebrate Paleontology contributes further evidence that whale evolution was not set along a unilinear path. The new research, conducted by Lisa Cooper, Hans Thewissen, and S.T. Hussain, focuses on the early part of whale evolution, from the time of Pakicetus to Ambulocetus. Contrary to what the popular illustrations present, there were many species of semi-aquatic whales living alongside one another for millions of years./p>

The focus of the new study is a particularly rich fossil site in the Kuldana
Formation in northern Pakistan which contains deposits between about 48 and 40 million years old. From bottom to top the formation records the incursion of marine environments into earlier freshwater habitats, and the diversity of early whales shifted along with these changes. (In fact, a number of early whales were first described from fossils found in this formation.) The level of most relevance to the new paper, however, is a limestone bed indicative of a freshwater habitat just a bit older than the earliest known fossils of Ambulocetus. Within these limestone layers are many teeth of early whales, including some that represent two new species of Pakicetus.


The restored, partial upper tooth row of a pakicetid whale made from a composite of teeth from multiple species. The top row is the view of the teeth looking down, and the bottom row is a view from the side. M=Molar, P=Premolar, C=Canine. (From Cooper et al., 2009)

When the genus Pakicetus was established in 1981 it was recognized that there were two species of the early whale. There was Pakicetus inachus, the remains of which had spurred the initial description, and Pakicetus attocki, which was represented by teeth originally attributed to another whale. Now Cooper and colleagues have added two more species on the basis of dental evidence; Pakicetus calcis and Pakicetus chittas. Of particular interest, though, is that the fragments of both species were found in geologically younger rocks than Pakicetus inachus and Pakicetus attocki. Rather than being found at the bottom of the Kuldana Formation they were found near the top, not far below the level from which Ambulocetus was exhumed, suggesting that several species of Pakicetus lived alongside each other for a long span of time.


A restoration of the skull of Remingtonocetus, a close relative of Attockicetus. From "Whale Origins as a Poster Child For Macroevolution" in BioScience.

But Pakicetus was not the only genus of early whale to be found in the limestone layers. The identification is still tentative, but there were also teeth from a relative of Kutchicetus called Attockicetus which was originally described by Thewissen and Hussain in 2000. The initial description was made on the basis of bones found in marine sediments of a slightly younger age than those that yielded Ambulocetus, so if the new assignment is correct is extends the temporal range of Attockicetus down the geological column. This means that it would have overlapped with the newly recognized Pakicetus species.


A stratigraphic map of the Kuldana Formation in Pakistan. The levels at which the remains of early whales have been found is indicated on the right. (From Cooper et al., 2009)

At first this might seem rather dull, but it has important implications for how we understand the evolution of early whales. Looking at the stratigraphic map of the Kuldana Formation from bottom to top it is apparent that Pakicetus inachus, Pakicetus attocki, and its close relatives Nalacetus and Icthyolestes all lived in freshwater habitats in the area at about the same time. There had obviously been a radiation of pakicetid whales early in the history of the group.

By the time the relatively more recent limestone layer was being laid down, however, the area was inhabited by a different assemblage of semi-aquatic whales. Not only were there two different species of Pakicetus (Pakicetus calcis and Pakicetus chittas), but also the long-snouted remingtonocetid Attockicetus. The fact that the remains of Ambulocetus are found not far above this layer suggests that this portion of the formation is from a time when early whales were diversifying from the earlier Pakicetus-type form. Some of those, such as Ambulocetus, can be slotted in to represent a transitional form between Pakicetus and living whales, while others (i.e. Attockicetus) cannot.

This pattern contrasts with the traditional image of constant change between forms along a straight line. Instead there appears to have been an early diversification of which few forms survived (primarily Pakicetus) but did not change very much. Yet this group of survivors formed the basis for a later radiation of forms that were adapted to spend more time in the water, with Pakicetus overlapping with some early members of this second radiation before going extinct. Unfortunately the record from this area is not yet complete enough to identify direct ancestors and descendants, but the overall pattern accords more closely with the pattern predicted by punctuated equilibrium than the popular view of whale evolution as a straightforward march to the sea.

Cooper, L., Thewissen, J., & Hussain, S. (2009). New Middle Eocene Archaeocetes (Cetacea:Mammalia) from the Kuldana Formation of Northern Pakistan Journal of Vertebrate Paleontology, 29 (4), 1289-1299 DOI: 10.1671/039.029.0423

More like this

Interesting - is there any idea about whether some of these diverse forms went extinct slowly or as a result of some dramatic mass extinction event that left only a few survivors to become modern cetaceans?

Hmm... for several reasons, it's difficult to unequivocally demonstrate that two fossil organisms lived side-by-side in the same environment (unless we're talking about gut residues, or a fossil fetus, or huge monospecific bonebeds etc.). I.e. just because two fossils occur in the same bed, doesn't necessarily mean they died in the same area, or even the same time, let alone lived in the same region; time averaging is a huge problem in terrestrial fluvial systems, and many taphonomic studies more or less indicate we're S.O.L. in cases like this (i.e. multitaxic assemblages).

There is also the possibility (actually, I can state it with certainty) that some of these are significantly time averaged, and that many of these fossils could be reworked. The lower fossiliferous units in the Kuldana are in conglomerates, in 'transgressive packages' - in sequence strat. terms, transgressive systems tracts are widely known for extensive erosion and cannibalization of underlying strata. The material described by Cooper et al. (2009) are from more transgressive strata towards the top of the Kuldana, so again - time averaging is probable here as well. The specimens are durable teeth and jaw fragments, which are readily transported and reworked.

Lastly... the archaeocete record is probably far too incomplete to tell one way or the other. For one, we all know orthogenesis doesn't happen (and to my knowledge no one is really pushing for that in cetacean evolution). Then again, is punctuated equilibrium an actual macroevolutionary pattern, or is it simply the effect of the incompleteness of the rock record? I haven't read much on the tempo and mode of invert evolution, so I can't vouch for them. However, the vertebrate fossil record doesn't lend itself very well towards tests of this idea because of its incompleteness and overal scarcity relative to calcareous invertebrates. It's blatantly obvious that a sampling/taphonomic/stratigraphic artifact of the fossil record can mistaken for punctuated equilibrium.

Well, I don't think that the fossil record will aver complete enough to find relationship of ancestors to progeny.

I mean, how can we know? We won't ever see this creatures mates, and then how can we infer ascendance relation?

Wnen we use cladistic methods to reconstruct the phylogenic tree of fossils, basically, you end up with a tree with all species at the end of a branch, never on a inner node of the tree.

So, you can't say intermediate : there's obviously animals bearing "transitional character" but no transitional animal.

Moreover, using the term transitional suppose that you have a trend, eg that the first "whales" which were not fully marine animals had the destiny to gave birth to fully marine form. And you'll agree that's a total nonsense. Evolution is a matter of immediate advantages, not a matter of trend.

So what happen to the poor Pakicetus here? is it a poor animal, meerly an intermediate, or is it an animal well adapted to its environment that, if it lived nowadays who won't ever suspect to give birth to anything different?

I'm not for the word "transitional". It's too confusing, sometimes just the term "evolution" : too much people thinks it mean "progress" like technology.

I prefer the cladistic methods which stopped to but fossils systematically as ancestors of younger groups : it can just be relatives, as you said. Since we can't prove scientifically they are real ancestor, we can only infer what ancestor should have looked like and say with caution that I-don't-know-which fossil looks like it. Because if you say "this fossil is probably the ancestor", there will be a journalist to draw a scale and but the whole bunch of fossil whales as ancestors and a blue whale at the end of it... Just like "the evolution", god, whatever, known from the beginning that there will be blue whales and do all the intermediates to reach this goal.

"Unfortunately the record from this area is not yet complete enough to identify direct ancestors and descendants"

Can you ever do this?

By anonymoose (not verified) on 16 Dec 2009 #permalink

Prawn, sorry - but that's nonsense. The lack of representation of 'nodes' in cladistics is due entirely to the way a tree is displayed (and the oft spoken assumption that you'll never find a node, which is unfortunately sometimes taken to mean we'll never find an ancestor in the record, which is total bull). If the fossil record is good enough with enough (good) fossils spaced stratigraphically, with a fair enough sample size, cladistics can be used in conjunction with other data to identify ancestors (i.e. within an anagenetic lineage). For example, take a lineage with taxon A,B,C,D, and E. They occur in successively younger strata. If cladistics shows that they diverge in that order (i.e. A first, than B,C,D, and E) then that is an example of cladistics identifying anagenesis as the most parsimonious interpretation. Anagenesis is a fundamentally simpler process than cladogenesis, and therefore *should* be the null anyway. In any event, we have to work with what we've got - if we can't accept for a second that any fossil is an intermediate or transitional between any others (and a living animal), you might as well just believe that evolution doesn't happen at all. Either way, it will be many years before the archaeocete record is good enough (in terms of stratigraphic sampling and total number of decent sample sizes therein) to test these ideas; however, the record is definitely excellent in terms of studying phylogeny (just not phyletic gradualism versus PE at this point).

Boesse : of course, if you have a anagenesis, it will be displayed as you say, but, with the data we have, we have no way to say that it was truly an anagenesis : how can you say that you have all the record, not only the dominant species, or the ones that lived in the right condition to be preserved as fossils? And I think that the post of Brian Switek prove that in the case of fossil whale, where're realy not in from of an anagenesis but in from of a wide radiation...

And to come back to the example you gave, the inner nodes that will links A,B,C,D,E in this pattern : (A(B(C(D,E)))) doesn't mean that A is the ancestor of all others, it may have lived at the same time than one or all the other, and therefore be a sister group. It could also have lived before all the other, but that don't necessary mean it was an ancestor.

So when you do a cladistic analysis, you never find ancestors, only parental relationship between sister groups, all the internal nodes are hypothetical ancestor. There was a real ancestor, but in cladistic, we discuss on hypothetical ones. In paleontology then, you will wonder if one species could be that ancestor. But to tell the truth, you won't ever prove it.

In the best case, you end up with the candidate ancestor (a real fossil) at the end of a branch with zero evolutionary step to the node of the hypothetical ancestor...

But, since the fossils aren't complete, since there's many things you don't know about the hypothetical ancestor, you cannot say that there's any evolutionary step on the branch between your fossil taxon and the reconstuction of the hypothetical ancestor.

So we can't reconstitute real ascendant relationships without doing supposition not supported by facts, that's what I meant.

So, you can't find an intermediate. Intermediate is a concept that shall not be used when you do science. If you do journalism, that's another story, but then, that's one of the origin of many misconception about evolution and science in general : we don't prove that something doesn't exist, in that sense in science, it's incorrect to think that you can have all the data.

So you can't prove that you have the complete fossil archive, and won't ever prove it, and you won't ever find all this info on the fossil to say with 100% confidence it's an ancestor, then the best attitude is to say it's probably close to the ancestor - no more - and certainly not an intermediate, an expression as correct as "living fossil".

Edit : in my previous comment, there's plenty mistakes, something related to my spell checker, jamming fingers, that it's late in the evening here or that english isn't my native language. Anyway, I write "in from" instead of "in front" twice, sorry. Now, I'd better go to bed,

I think stratocladistics enables you in principle to identify direct ancestors. But I think this remains a minority approach. In regular cladistics, you can't do this. It might have been Hennig who first argued this. But I'm out of my depth here and could be wrong.

By anonymoose (not verified) on 16 Dec 2009 #permalink

What continues to puzzle me is why we still make such a big deal about any new evidence showing that a previously apparently unilinear sequence of descent was in actuality a bush with multiple adaptive radiations, just like every other lineage about which we have sufficient data, as if it were some shocking, ground-breaking, overthrowing of previously assumed knowledge.

I mean, at this stage of the game, we really should be taking bushiness as the default null hypothesis for every evolutionary lineage, with unilinearity being solely an artifact of incomplete preservation, and downright be expecting evidence for branching lineages to be found. And any representation resembling a unilinear descent should be assumed to be missing the side branches (either because we don't have the fossils or other evidence, or the representation has been deliberately simplified with the removal of these branches, for the sake of illustrating a point with greater clarity)

In fact, I would regard conclusive evidence for a truly unilinear line of descent, without any branching, to be a solid indication of divine intervention.

Arranging species in a line is arguing patristic speciation. I would be sceptical of this and would want continuous evidence much better than can be expected from scattered fossils here and there.

By Jim Thomerson (not verified) on 16 Dec 2009 #permalink

By the way folks, I'm arguing for anagenetic lineages on a "subgeneric" level - i.e. species level - NOT orthogenesis. We all know that doesn't happen. I do agree with Amphiox, though - I'm not sure why folks make such a big deal about that, and "cockroach" taxa - although I read the paper this morning, and nothing regarding that sorta stuff was mentioned (it focused on pakicetid and ambulocetid dental morphology).

Prawn et al. - if ABCDE didn't fall out stratigraphically or overlapped, then that's a simple test of anagenesis =) Read what I said again. The cladistic pattern is one half of it - you need the clear stratigraphic succession as well to demonstrate it as the most parsimonious hypothesis. It's a testable hypothesis. It's no less valid than cladogenesis - I say it should be the null because it is fundamentally (and thermodynamically, to put it that way) simpler - i.e. no lineage splitting.

My problem with amphiox' comment RE: never expecting to find a lineage in the rock record, and expecting branching to always happen in the record. So, every species, EVER, has branched? Lineages don't exist, at all? ever? Our own species has existed as a non-branching lineage for much of the latter Pleistocene; we haven't undergone constant cladogenesis. If you argue the point that lineages don't exist, soon every generation is a different 'species' in that sense. Anyway, read Gingerich's paleobio paper on stasis in Eocene mammals from Wyoming. Example of lineages in the fossil record, right there (yes, with 'N'fairly friggin high up there).

Prawn - You say I can't 'prove' anagenesis - you're damned right. But I wouldn't be trying to - you can't prove it just as much as you can't prove cladogenesis - you can only identify it as the most parsimonious conclusion. Nothing's ever really 'proven' in that sense anyway.

Many of your folks ideas are undoubtedly born out of the crappy state of the vertebrate fossil record (which I alluded to earlier, RE: not being able to actually assess whether any of these taxa co-occurred, and may be reworked). Vertebrates are not the only fossils, guys! People who work on invertebrate fossils don't have some mysterious hang-up on dangeous subject matter like 'lineages', 'anagenesis' and (lo and behold) simpler, run of the mill-type evolutionary processes. In fact, they routinely identify ancestor-descendant relationships and lineages (and cladogenetic events, mind you) in the rock record - and shit, they use those methods for stratigraphic correlation! And, better yet, to find oil (i.e. forams, radiolaria, diatoms, etc.)! For one, the invertebrate record is good enough to do that (i.e. molluscs and microfossils), while most of the vertebrate record is not. However, certain parts (i.e. rodents, horses, sharks, possibly turtles) are probably good enough to lend themselves to stratocladistic analyses and determinations of lineages.

Anyway, sorry for the long comment.

> Interesting - is there any idea about whether some of these
> diverse forms went extinct slowly or as a result of some
> dramatic mass extinction event that left only a few
> survivors to become modern cetaceans?

The climatic change from "hothouse" to "icehouse" conditions during the late Palaeogene/early Neogene must have hit many species hard - especially those who dwelled in warm, shallow coastal waters. The more terrestrial and cursorial forms, like *Pakicetus*, also had a lot of competition in their predatory or omnivorous niches from triisodonts, mesonychians, arctocyonids, hyaenodonts, oxyaenids, ziphodont crocs, gastornithid birds (although the latter might have been herbivorous), varanid lizards and, later in the Eocene, also from amphicyonids, nimravids and entelodonts, so it was perhaps inevitable that some of those clades had to go.

Prawn et al. - if ABCDE didn't fall out stratigraphically or overlapped, then that's a simple test of anagenesis =) Read what I said again. The cladistic pattern is one half of it - you need the clear stratigraphic succession as well to demonstrate it as the most parsimonious hypothesis. It's a testable hypothesis. It's no less valid than cladogenesis - I say it should be the null because it is fundamentally (and thermodynamically, to put it that way) simpler - i.e. no lineage splitting.

Suppose we find (A(B(C(D,E)))) in our cladistic analysis, that A preceeded B preceded C and so on stratigraphically, etc, but also that C has a bunch of autapomorphies. If ABCDE is an anagenetic lineage, there thus was a bunch of reversals between C and D - how many such reversals can we have and still consider anagenesis a simpler explanation than that C represents a side-branch?

(That's a serious question - I've heard it suggested that any apparent reversals at all should make us assume cladogenesis, but that seems dogmatic. Reversals do happen, and we can't know if the autapomorphy was even universal within C. But equally, we know that branching is common, and there must be a point where one branching is likelier than many reversals.)

By Andreas Johansson (not verified) on 17 Dec 2009 #permalink

Andreas - that's an excellent question. Again, something extremely easy to test cladistically; it all boils down to parsimony.

Johannes - several of Ewan Fordyce's papers have addressed Paleogene cooling and cetacean evolution - but if I understand that correctly, that happens much later, closer to the Eocene/Oligocene boundary - by the late Eocene, pakicetids and ambulocetids are largely extinct, and we're left with basilosaurids (and possibly some protocetid stragglers; I can't remember). So you're right about it affecting early whale evolution, but it affects later diverging archaeocetes, and has been blamed before for the radiation of Neoceti.

that's an excellent question. Again, something extremely easy to test cladistically; it all boils down to parsimony.

But how exactly does one weight cladogenesis against reversal?

By Andreas Johansson (not verified) on 17 Dec 2009 #permalink