Over at Evolving Thoughts, John Wilkins has a post that criticizes a recently-published journal article. Normally, I agree with John - in fact, if it's true that the best measure of someone's intelligence is how often their views match yours, then John Wilkins is an absolute genius. But even Einstein had off-days, and (again, based on the agreement standard) I think this might have been one of John's.
The article in question, by paleontologists Sarda Sahney and Michael Benton, examines how long it took for ecosystems to recover after the end-Permian extinction. The dinosaurs weren't around then, so the end-Permian doesn't usually get the attention that the end-Cretaceous does, but it was by far a much more significant event. By some estimates, more than 95% of all animal species went extinct at that time. John's main complaint is with the way paleontologists compute estimates like that, and I'll explain why his objections are a bit on the idealistic side later on. First, let's take a look at what the authors did, what they found, and why their results are way cool.
When scientists study mass extinctions, they are usually interested in one of two things: what happened to cause the extinction, and what happened after the extinction. Both of those questions are important, not just because they help us understand what things were like millions of years ago, but also because they can help us better understand the potential consequences of some of our actions right now. In this study, Sahney and Benton are focusing on the second question. They want to know how long it took for new species to evolve and re-establish the level of diversity that was seen in terrestrial tetrapod ecosystems before the Permian extinction.
There are a number of different ways to measure diversity. One of the most common is to look at alpha-level diversity. Assessing the alpha diversity in some location or at some time is a comparatively simple process. You count the number of taxa that are present, and that's the alpha diversity. The authors did this for terrestrial tetrapods during the period of time that they were interested in. If you just look at that measure, though, it doesn't look like it took terrestrial tetrapods all that long to recover from the extinction event. The problem is that alpha diversity really isn't a very good measure of what's actually going on in an ecosystem. All it tells you is how many species there are. It doesn't tell you what roles the species are filling, and doesn't tell you if the ecosystem is dominated by one or two really, really common species. If you want to get a better idea of what's happening in ecosystems, you need to look deeper.
And that's just what Sahney and Benton did. They used the body size and inferred diet of the tetrapods that were found in a number of well-sampled fossil localities to determine the number of ecological guilds that were filled over the same time period.
The guilds are defined by body size and inferred diet. Error bars cannot be applied as the number of occupied guilds is not a mean number, rather a cumulative count of the guilds filled in each stage. Size of the animals is defined as small (with snout-vent length (SVL) less than 150mm), medium (SVL from 150mm to 1.5m) and large (SVL greater than 1.5m). Geological stages (Gradstein & Ogg 2004) are as follows: Ar, Artinskian; K, Kungurian; R, Roadian; W, Wordian; Ca, Capitanian; Wu, Wuchiapingian; Ch, Changhsingian; I, Induan; O, Olenekian; An, Anisian; L, Ladinian; Cr, Carnian. (Figure is Figure 3c of Sahney & Benton 2008, caption is Figure 3c caption with addition of geological stage abbreviations from their Figure 1.)
Using this measure, you can see that recovery from the extinction event was a long, drawn out process. It took about 15 million years for diversity to reach pre-extinction levels, and it was a lot longer before the situation stabilizes. To put it into human terms, this suggests that if we manage to totally and completely trash the planet, our grandkids probably aren't going to be able to fix it.
If you haven't guessed yet, I think the paper is pretty cool. It's an interesting new look at the dynamics of evolution and extinction over long periods of time, during one of the most dramatic periods of biological upheaval our planet's seen. The authors put a lot of effort and creativity into their study, and it will be very interesting to see if their methods can be applied to other periods of time. So why isn't Wilkins as enthusiastic about this paper as I am?
The problem that he sees with this study involves the taxonomic level that the authors were looking at. They didn't make their estimates by looking at species, or even genera. Instead, they looked at the number of families that were present. He doesn't think that's kosher:
Sahney and Benton use a Linnaean rank as a surrogate for diversity of species, or rather, two ranks: orders and families. The problem with this is that it is unclear how objective such ranks are. Linnaean ranks were even by Linneaus' own standards, fairly arbitrary - in fact Linnaeus, who had five ranks only, said they were merely an aid to the student and the collector (and a very good aid they were too).
But apart - arguably - from species, none of these ranks represented anything real. Oh, Linnaeus thought kingdom did, but recent work shows us that the kingdoms are just arbitrary bits of the phylogenetic tree marked out because it suits us to do so. Plantae might be real, but Animalia seems to be artificial, as Fungi ought to be included with it.
The particular rank used by Sahney and Benton - family - isn't even one of the original Linnaean ranks. It was a term of Michel Adanson, a younger contemporary of Linnaeus, and didn't get included into Linnaean ranks until 1871. So what does it indicate in the natural world?
I really can't argue with any of that. He's right. A biological family is a human construction. There is absolutely no objective definition of a "family", and there's no non-arbitrary method for determining what constitutes a "family". "Families" have no biological reality. But there are a couple of important things that he doesn't take into consideration. For one, families may be bad in this kind of study, but if so it's in a very Churchillian way - they're the worst except for all of the other possibilities. For another, Sahney and Benton's methodology goes a long way toward addressing the problems he outlined.
When authors look at mass extinctions in the past, they're (obviously) looking at the fossil record. The problem is that the fossil record is massively incomplete by any standard. The vast majority of organisms that have lived and died on this planet have done so without leaving even the faintest hint of a suggestion of a smidgen of a trace that they were ever here. That's particularly true when it comes to terrestrial tetrapods. Large animals (and by fossil standards anything bigger than shoe box is large) do not fossilize well, and things that live on dry land do not fossilize well. The fossil record for large animals that lived on dry land is spectacularly incomplete.
The result of the incompleteness in the fossil record is that scientists simply can't use species or genera for these sorts of studies. There are so many cases where our entire record of a species or genus consists of a single specimen that any attempt to use those levels to plot diversity would result in incredible confusion. So instead, scientists use families as the best available proxy for what was happening at the other levels. That doesn't mean that it's a perfect proxy, or even that it's a very good one. It's just what there is.
By looking at ecological roles instead of taxonomic classification, Sahney and Benton were able to get around these problems - which is one of the best things about their paper. They plotted their guild measures based on the presence or absence of representative animals in the fossil record. This provides a measure that's independent from how the animals are classified. The results that they obtained from assessing community structure are reasonably compatible with the results obtained by looking at the taxonomy. That's great, because it indicates that, poor as it may be, using families to assess diversity over time might not be a complete disaster.
By no means, though, should you take my word for any of this. Go look at the paper, and at John's article, and at what Greg Laden has to say about it, and at Sarda Sahney's blog. While your at it, you might also want to take a look at researchblogging.org, where you'll find many other bloggers writing about things that they've read in the peer-reviewed literature.
Sahney, S., Benton, M.J. (2008). Recovery from the most profound mass extinction of all time. Proceedings of the Royal Society B: Biological Sciences, -1(1), -1--1. DOI: 10.1098/rspb.2007.1370
It seems to me that if the families are formed by the same subjective interpretations, then they possibly may represent something objective, but I really doubt it. While I have immense respect for the cognitive capacities of taxonomists who have immersed themselves in the study of their organisms, inference based on these subjective ranks is in effect only making inference from the taxonomists' properties, not the organisms as such. Any way you cut it, it's an illicit inference.
We think there may be an extinction pulse here, and by trying to work out the ecological communities - also, incidentally not objectively defined - we may try to work out how great, but all the figures of that graph are arbitrary.
Something that I forgot until just now, but which I think may be important here is the nested nature of the taxonomic hierarchy.
Families are arbitrary and artificial constructs. However, the extinction of a family by necessity requires the extinction of at least one species, and the origination of a family requires the origination of at least one species. So measuring changes in diversity at the family level over a period of time should be equivalent to measuring a sampling of the changes in diversity at the species level over the same period. It's a non-random sample, though, and that presents another (if slightly different) set of problems.
The concept of ecological guild and what it means to "fill" a guild is crucial to understanding the article and your comments on it. I found 2 definitions of guild: (Wikipedia) "Guilds are groups of organisms sharing a similar way of life... They are classified according to how they acquire their nutrients, their state of mobility, and their mode of feeding." And this: (ecology article Univ. of Texas) Guilds are "groups of functionally similar species in a community, such as foliage-gleaning insectivorous birds...Techniques for objectively defining a guild remain in their infancy.." This brings up a large number of questions. If the guild is defined by the nutritional niche of its members, then in a mass extinction many guilds would disappear when their food source dies out. The former guild's members might also become extinct or they might become members of other guilds as they adapt to a changed habitat. It seems to me that you would need a fairly detailed picture of the initial ecology and its changes at multiple trophic levels to say anything meaningful about guild membership and structure. How do you know how much you are missing due to sampling error as the species diversity recovers? How do you compare the guild membership of trilobite predators without trilobites? Did all of their predators die out? Do you wait however many million years until the trilobite niche has been imperfectly filled by something else?
The recovery from an extinction event is certainly reliant on surviving species and their evolutionary descendents. How do guilds relate to evolutionary descent (cladistics)? The use of "family" and guild membership as criteria of loss and recovery is very confusing. When I try to grasp the diversification and multiplication of the real creatures through the abstraction of guilds and families, I feel like I am grasping sand.