Extinction Matters

Adrian Currie writes...

Does Paleobiology have laws? If you’d asked me a week ago I would have laughed. Now, I’m still laughing a little, but I’m a little more open to the idea. Maybe.

One of the delights of my job is attending scientists’ reading groups, lab meetings and journal clubs. I get to watch and interact with scientists in one of their natural habitats: kicking around a paper, figuring out how it works (and doesn’t), pouring over tables, crossing our eyes at figures and equations, gossiping, and so forth. This is both a great privilege and, from a philosophy of science perspective, money for jam. I get to learn about new scientific developments, from the horse’s mouth, simply by hanging out with a bunch of lovely people who know much more about the things than I do. And I’ve been very lucky: by and large scientists been very welcoming and patient with me.

Anyway, last week Cambridge’s paleontology journal club read a paper, with a title of pure, uncut philosophical catnip: Charles R Marshall’s Five palaeobiological laws needed to understand the evolution of the living biota, just this minute published in a new journal (Nature’s ecology & evolution). Seeing that title, I thought gee, there’s got to be a blog post in that! After all, philosophers of biology have long been pretty sceptical of the existence of laws in biology, let alone paleontology. And indeed, many philosophers of science are pretty dismissive of all this ‘law’ talk anywho (witness Woodward being pretty forceful about this, although Robert Brandon and Dan McShea probably think otherwise…)

So, I was expecting to write something about laws, describing some philosophical thinking on what laws were (probably touching on Sandy Mitchell and Marc Lange’s work), and asking whether Marshall has identified anything like what I might properly call a ‘law’, all the while being a little snooty about the whole thing. But upon reading the paper, I saw that it was much more interesting than I thought. The fun isn’t really in paleobiological laws at all, rather the action is in the relationship between fossil data and living critters, and how important extinction is for understanding life.

In this post, I’ll describe the paper and my thoughts on it. Then, I’ll briefly report back on the discussions in the journal club.

Marshall’s motivation is one which has underwritten my work for a while now: the importance of the fossil record in understanding life. He points out that although everyone recognises that differential survival—who lives and who dies—matters crucially for evolution at a population level,

On longer timescales, however, although it is well recognized that there has been extensive termination of unique morphologies and lineages, study of the living biota alone offers relatively little when trying to determine the importance of these extinctions, given that among the living we only have direct data on the survivors.

Happily, the fossil record might just have the information we need to answer these crucial questions about extinction.

Let’s put things a little more dramatically. If you’re going to ask life’s big questions: about its nature, its patterns, and the forces which shape it, you gotta care about palaeontology.

After all, life today is super diverse right? But it’s only a tiny fraction of the total life that has existed. All of those echidnas, ants, naked mole-rats, lung-fish, sequoia, and so forth, are squeezed into the tiniest fragment of time. And there’s reason to think that time-slice isn’t particularly representative. Moreover, some of life’s major events are simply invisible from a purely presentist or neontological perspective. If we only had extant animals to go on, we’d have no hint of, to take one example, the enormously successful, diverse, lineage out of which the birds emerged.

If we want to know about major trends and regularities at an evolutionary scale, then, we need a paleobiological perspective.

Marshall’s argument for this view focuses on rates of extinction. The background thought, in essence, is that from the perspective of molecular phylogenetics, which (of course) is more-or-less limited to our current time-slice, a species’ extinction leads to its erasure. If, then, you want to consider how life’s shape has changed—in this case how rates of diversification have shifted across time—you need to have a sense of extinction rates and speciation rates. The thinking is similar to models in population ecology, where you need a sense of birth rates and death rates to understand how a population will change over time. Marshall’s complaint is that molecular studies which have tried to look at diversification get things wrong because they are unable to ground their speculations about rates of speciation (diversity increases) and extinction (diversity decreases) in meaningful empirical information.

Before describing the laws themselves, I want to note a little bafflement at the thought that people would use molecular phylogenetics to study radiations and so forth in the first place (at least at such large scales). I thought that phylogenetics told us about, first, the ancestral relations between lineages (so, whether humans are more related to chimpanzees or orang-utans) and, if we can get our clock ticking right, the rough timing of the divergences between those lineages (that, say, humans and chimpanzees had a common ancestor around 6 million years ago). From that information we can start making inferences about the nature of those ancestors (say, did the common ancestor of humans and chimpanzees have a chimpanzee-like social structure). And that is important for us generating explanations of the evolution of particular lineages (for instance, if I want to explain hominid evolution, I need to include a story about how we get from a chimpanzee-like social system to a human-like social system). Although on small scales, where we’re not likely to lose information due to extinctions, molecular data is likely fine, it never quite occurred to me that we could use extant biota alone to generates rates of speciation and radiations over evolutionary time. Anyway, enough ignorant bafflement from me, let’s sketch the ‘laws’ (I’ll avoid the equations, if you want the juicy specifics check out the paper…).

Okay, the first law is simply that lineages go extinct.

Woah, really...?

Woah, really...?

More seriously, though, the point of the first law is that it demands that an extinction rate be used: basically, each lineage has a probability of going extinct above 0.

The second law states that because lineages are short lived, current biota are massively outnumbered by past biota. This can be represented by taking species longevity to be: 1/extinction rate. Marshall briefly examines work which reconstructs the fossil history of mammals as an example of how to estimate species longevity.

By the third law, the origination rate of lineages (I presume these are branching events) is roughly equal to the extinction rate—at least on average, ignoring early expansion (so far as I can tell). So, the one variable (extinction rate) determines both the rate at which lineages are generated, and the rate at which they are pruned. Here, Marshall claims “strong support from the fossil record, where it has long been noted that, even for living groups, measured long-term origination rates only slightly exceed long-term extinction rates.”

The fourth law tells us that variability in species richness across evolutionary time is due to shifting extinction and generation rates. So far as I can tell, this is basically determined by the second and third law, although Marshall also points to fossil reconstruction work which suggests that it also has empirical content.

The fifth law states that extinction erases history—as I mentioned above, from an extant, molecular perspective if you go extinct, you disappear. It’s only thanks to the fossil record that we have any inkling of T. rex and her kin.

Why does all this matter? I take it the thought is that if you’re going to use molecular phylogenetic methods to estimate species diversity over time, you’d better make sure that you include extinction (by law 1), and include that origination events and extinction events are roughly equal (law 3). Moreover, you’d better let those rates shift as well—that’s the source of shifts in diversity (law 4), and hey, you’d better determine the extinction rate using fossils, because there were masses of lineages in the past (law 2) and their signal has been erased from the molecular record (law 5).

To sum up, if molecular phylogenetics is going to inform us about macro-evolutionary diversity, extinction matters. And understanding extinction means we need to understand the fossil record.

Marshall sells his laws as a way of taking the fossil record seriously when understanding extant biota, and thus we might imagine that would involve integrating fossil and molecular data. However, I don’t think this is the right characterization of the presented relationship. Distinguish between ‘integration’ and ‘calibration’. The former involves data from two sources being incorporated together to infer the result; the latter involves one source of data being relied on to set a few variables which are then used to analyse the other. Integration uses all or most of the available evidence, while calibration does not.

An example of phylogenetic integration is these new-fangled techniques paleontologists are using to generate phylogenies including both fossil and molecular characters. I won't get into the details here, but *very* roughly speaking, characters from both sources of data are used to construct various phylogenies, typically in a Bayesian framework which allows necessary assumptions about the relationship between molecular and morphological characters to be built into priors. An example of phylogenetic calibration is the use of the appearance of fossil lineages in strata to set the timing of molecular clocks. To estimate time differences in molecular phylogenies we need some way of working out the pace at which genes change. In the right circumstances, we can compare the results of our guesstimates of molecular rates to the (more) solid dates provided by fossils in strata. Calibration is very different from integration: in the former, although fossil data gets to speak, it's only briefly.

Given the distinction between calibration and integration, we can generate a continuum between phylogenetic techniques which rely on fossils maximally (as in ‘pure’ morphological phylogenetics) and those which rely on them minimally (as in molecular phylogenetics without fossil-calibrated clocks). In between, we have morphological studies which calibrate on molecular information (does such a thing exist?), true cases of integration, and molecular phylogenetics with morphological calibration.

So, where do Marshall’s ‘laws’ fit on this continuum? I think it’s pretty clear that it is a case of morphological calibration. We are provided with an equation which allows us to incorporate extinction and speciation rates into molecular data. The variables in the equation, such as extinction rates, are set on the basis of fossil data. This is strongly analogous to using fossils in strata to calibrate molecular clocks.

I think this matters—negatively—on the basis of two points. First, on the assumption that we want to maximize the amount of data we’re using, if fossil data’s only role is to calibrate the extinction rate on our dispersal models, then we miss out a whole bunch of information. Second, the motivation of the piece, as I put it above, is that the paleontological record matters: it is a crucial source for understanding macroevolutionary patterns and processes because it is the only source of information we have which is at the right scale. Merely using it to calibrate seems to do insufficient homage to that mattering (perhaps a way to get around all of this is to use Marshall’s equations on a properly integrated phylogenetic approach).

Anyway, I think the paper’s pretty interesting, but it isn’t pro-fossil enough for my taste… Let’s see what happens at the journal club…


Right, first off: everyone at the club agreed to my passing on a few things under the Chatham House Rule, and although there were a fair few differences in people’s opinions about some of the details of the paper, there was a general approval of it. So that’s nice. (and, in this group, rare!)

Unsurprisingly, there was a bit of discussion of the ‘are these laws? What are laws anyway?’ variety. It’s pretty clear that Marshall’s laws aren’t laws with forces, as say Newton’s are, although there was a suggestion that they could be thought of as analogous to statistical mechanics. I’m not so sure of that (statistical mechanics can make kinds of predictions and explanations which I’m not so convinced Marshall’s laws can). I’m still inclined to think that the best analogy is to the ‘laws’ of population ecology, or perhaps economics (and – I’m still pretty convinced that the ‘law’ talk is a red herring, concept & epistemology-wise).

Law 3 led to a bunch of headaches: in particular it was rather unclear whether the law is, as it were, a priori (derived from laws 1 and 2 perhaps), or whether it was empirical—extrapolated from patterns in the fossil record—or somewhere somehow in between. In one sense, average origination and extinction should be equal on average across the lifespan of a clade, assuming that it starts with 1 species and ends with 1 species. But this isn’t what the law claims: rather, is seems to be saying that, for most spans of time (discounting early radiations and final extinctions) origination and extinction even out. The law is also explicitly approximate and caveated (us philosophers would say ceteris paribus) which certainly suggests something not so a priori. Generally speaking, it’s a bit unclear whether these laws are stipulated, and then filled in with empirical data; or whether the relationships they capture are themselves empirical.

One interesting difference between my reaction and some of the others was that I took a narrow reading of the laws: that they were, in effect, some equations that neontologists could use to ground their inferences from molecular phylogenies to diversity rates in fossil data. Although there was agreement that the paper was really only interested in helping ‘our neontological brethren’, others took the view that the laws had (and were intended to have) a much wider scope of application. This helped explain the presence of law 5, which didn’t look to me as really fitting with the other laws, and was rather a claim about our epistemic situation vis-à-vis extinct lineages. Relating to law 5, the discussion’s convinced me to be much more suspicious of the kind of ‘common ancestor’ inferencing that goes on a fair bit in neontology (and I’ve been pretty optimistic about generally speaking, especially here and here). Indeed: I think Richard Javier Stephenson’s recent post might also be relevant here. Watch this space: I reckon a relook at homologous inferences given this kind of objection might be on the cards…

Finally, it was pointed out that this paper actually underwrites an important research program in paleontology. It’s long been the case, both in paleontology and neontology, that people have bemoaned the lack of researchers at the natural-history coal-face, identifying lineages and their characters, new species, and so forth. Everyone agrees such work is important but, it’s a bit, well…

If you don't get it ask Rutherford...

If you don't get it ask Rutherford...

However, if Marshall’s right, then if we’re going to use molecular data to understand rates of diversity over evolutionary time—something close to the heart of evolutionary theory and thinking—then we really need to understand the patterns of diversity in particular lineages in the fossil record. Only by doing this are we able to get a handle on actual extinction rates, and thus put empirical clothes on Marshall’s laws. And that job requires careful fossil stamp collecting by those systematist heroes.

The big news, then, is that not only does extinction matter, but so too do fossils.

Thanks to the folks at the Journal Club, and to Derek and Leonard for reassuring me this wasn't a completely silly idea for a blog post.