Real Audio Recording   

Today I'm going to discuss the big picture with respect to the issue of the terminal Pleistocene mass extinction event in North America. And when I say "big picture," I mean it in three different ways. First of all, I'm going to discuss the very large spatial scale of the continent of North America -- not just one place or another, but as much of it as we can discuss. Secondly, I'm going to discuss the very long temporal scale of the Cenozoic Era, all 65 million years of it. And, finally, I'm going to discuss all of the mammals -- not just a few species, not just the victims of the extinction, but also the survivors.

Now, because the preceding speakers have set up these questions very well, and because there are a lot of subsidiary arguments you could make, and a lot of different things you could do with the data, and I don't have much time, I'm going to focus on two specific points. And this is going to involve two different data sets and two different methods of analysis. So, in the first half of the talk I'm going to discuss extinction intensity in the Late Pleistocene, and I'm going to try to answer the question of whether the intensity, and also the selectivity of that extinction, was unusual in the context of the entire Cenozoic -- the big picture.

In the second half of the talk, which will be much briefer, I'm going to discuss the issue not of extinction per se, but of organization of mammal communities. And this might seem like an abrupt shift, but it's relevant. Because although the overkill hypothesis of Paul Martin and others doesn't speak too much to the question of how surviving mammal communities should be organized, other hypotheses involving climate change and habitat change do make specific predictions. What they say is that the changes in habitat and climate should cause so much trouble for all of the mammals that you should see an imprint of that, not just in terms of geographic ranges going to zero and species becoming extinct, but other geographic ranges decreasing not to zero, but enough so that you see in the past associations of species that you no longer see in the present, due to range contraction.

So first I'm just going to summarize some points that I believe most of the speakers this morning can agree on. I might not be right -- it's just a guess. This is just to get out of the way possible other questions that might come up. I'll go through this very quickly. We've already seen that the North American event was coincident with both the deglaciation that was very rapid and extreme and with the first appearance of the Clovis hunting technology. That might not mean first appearance of humans, but it certainly is an interesting archeological event. We know that the extinction was probably very rapid -- perhaps took a few hundred years, perhaps a little more than a thousand. We know that there was intense selectivity targeted at large herbivores -- the larger, the more extinction. We know that the event occurred everywhere in the continent, from all the way in the north to Florida, to Arizona, and that a similar event happened all the way down through the Americas, all the way to Patagonia. The event in South America was similar in terms of selectivity, timing, intensity. Despite the dramatic differences in the taxonomic composition of the fauna and the dramatic differences in the habitat and climate in South America.

And the other point that's been raised by Paul Martin already is that this deglaciation event was not unique in the Late Pleistocene. The Late Pleistocene, in the broad sense -- meaning the last million years of the Pleistocene -- is an era of very intense glacial cycles, and the most extreme events in those cycles are deglaciations. And, repeatedly, roughly every 100,000 years, there's been an intense and rapid deglaciation event.

So, to me, at least, this last point is a fatal problem for those who believe that climate and habitat change, by itself, was responsible for the extinction event. If that's the case, they have to explain how the biota of the Western Hemisphere survived all of these preceding cycles without a detectable event, despite the fact that one would expect natural selection to increase the resistance of those faunas to climate change and habitat change. That's not to say that climate and habitat change didn't play any role -- just that I don't believe they played the sole role in the event.

But that doesn't settle the issue by itself. We still have these two particular hypotheses to deal with: Was the extinction event a natural type of an event? And did the extinction event accompany reorganization of the surviving mammal community?

This is to give you an idea of just how dramatically the mammal biota of North America has changed during the Cenozoic. This is a reconstruction of a Paleocene mammal biota, about 16 million years ago, dominated entirely by small mammals, belonging mostly to entirely extinct taxonomic groups. By the middle of the Eocene, about 45 million years ago, or 50 million years ago, you do have large mammals in the biota; you do have dedicated carnivores -- but many of these large herbivores and carnivores again belong to extinct taxonomic groups, like the vintathere in the middle. And it's only by the Middle Miocene -- this is between about 10 and 15 million years ago, in this reconstruction -- that you have members of the major groups that were around in the Pleistocene: Proboscidians, like those gompotheres in the corner; lots of horse species, like these; lots of endemic artiodactyls; pronghorns, were the major or endemic, artiodactyl group to make it to the Pleistocene, in addition to peccaries. There were rhinos in the Miocene, but they didn't survive into the Plio-Pleistocene.

So you start, at this point, to have a recognizable modern fauna, and the question is: Did all of this change occur because of episodic major extinction events causing replacements of the fauna? As you might expect, these large extinction events are natural -- so, to answer that question, I'm going to go to a database that I've been working on for a decade now. It involves what are called "faunal lists," which are inventories of the species found in particular fossil localities. There are about 4,000 of these. The data are based on about 2,400 references. I didn't have an assistant, so I had a lot of bleary eyes from reading papers. These statistics aren't particularly relevant, except to indicate that I've worked hard to get the taxonomy right by identifying invalid genera and species, and by removing questionable identifications, and also to indicate that we have a lot of time control, and indicate that I'm an Internet geek -- the database is available on a website.

This is to give you a feel for the data. This is a series of paleogeographic maps with not very reliable coastlines, indicating the geographic location of the faunas. This is the set of Cretaceous faunas, concentrated in the western interior -- and, as I move through these slides quickly, you'll note that the cloud of points stays mostly in the western interior, but eventually expands to cover the whole continent. You'll also note that, in some time slices -- and these time slices are roughly 10 million years apiece -- there's a lot more sampling than others. Here's the Paleocene, right after the K-T mass extinction, before the arrival of a new wave of groups in the Eocene. The early Eocene -- very, very good sampling in a very, very small area. Late Eocene -- the cloud's starting to expand. The Oligocene -- same type of western interior sampling. Early and Middle Miocene: Now you're finally picking up sampling on the Gulf Coast ... but sampling's poor. Whereas, in the Late Miocene, sampling is very good, particularly in the western half of North America, and you do have east coast sampling. And, finally, the Plio-Pleistocene in the last 5 million years. These are data exclusive of the Last Glacial and of the Holocene. So even if you ignore the very large number of relatively young localities that straddle the extinction event, we still have very good sampling across the continent.

So what do we do with all these faunal lists? Well, it's a bit complicated, and I don't have much time to discuss all the details. But the basic idea is that we want to get these lists into a temporal sequence, so that we can infer from the lists a diversity pattern, including extinction rates. And the way we do this is with a multivariant statistical method called "appearance event ordination." The method essentially takes the faunal lists and shuffles them, until it gets what's called a "parsimonious sequence" of first and last appearances of species and genera. These are originations and extinctions, essentially. Once you have a sequence going from oldest to youngest, that sequence defines a series of age ranges, which are like life spans of species and genera. Once you've obtained a parsimonious or a robust sequence, you can number it from oldest to youngest, and then the numbers can be used to give the faunal lists their own numbers, based on what are called the "concurrent range zones" of the species that are included in the lists.

Now, fortunately, many of these faunal lists are associated with completely independent geochronological age estimates, based on methods such as paleomatic fission track dating, argon-argon dating, potassium-argon dating, etc. And because those geochronologic estimates on the Y-axis are independent, we can use them to confirm that the ordination of the 4,000 faunal lists is really telling us about time, and not about something else. So there's a good strong monotonic relationship. And now we can use this relationship to back-calibrate the appearance event sequence itself, and therefore the age ranges.

And now we have the fixings for a diversity curve, but we're missing a couple of things. One of them is that we've analyzed genera and species, but we don't really care about the genera per se. We want to know about the species. Fortunately, the genera tell us something about the species that they include. If these species are all in the same genus, and there's a time when there's no species in that genus present, we can fill in the gap -- we can range through it, essentially. So we can use the genus-level data to fill in the gaps in the species-level data.

And now we've got our diversity curve. Starting in the Cretaceous, where diversity's low, exponential diversification. This is a logistic growth pattern. An equilibrium through the rest of the Cenozoic, with a lot of up and down. What that slide shows is a massive spike in the Pleistocene. And this is not due to the last 100,000 years, or 70,000 years of sampling -- it's due to sampling before that in the Pleistocene. We see a big increase in apparent diversity.

But that increase is not a real one -- it's due strictly to the number of faunal lists. More or less, more species, higher diversity. So we need to get rid of that signal somehow. And in the extremely aesthetic slide that you might have seen next, there would have been a representation of the sampling intensity through the Cenozoic, showing that sampling intensity rises and falls, and rises and falls, and then rises again in the Pleistocene. And these changes are on order of magnitude large -- so they're quite serious -- and that does present a problem, in that we can't completely trust the raw diversity curve we've obtained. Fortunately, it's not too difficult to remove that sampling signal using a method called rarefaction, that's standard in ecology. And all you're doing with rarefaction is throwing out lists in each time interval until you have a constant level of sampling in each interval.

And now we have pretty good confidence that sampling is not creating any of the patterns. And we also have confidence that the diversity pattern is not an artifact of the time scale, because we've been able to impose equally spaced one-million-year events on this diversity curve. In the brief time you had to see that purple slide, you saw a lot of points, and those points were equally spaced at a million years. So we've got a robust time scale; we've got control for sampling; we are making use of all the data that we could possibly obtain; we're looking at genera and species -- all the faunal lists. Now we've got some data that we can trust.

So here's the correct curve, which shows all the same essential patterns: exponential growth, equilibrium, some up and down that's meaningful. Here's this end-Pleistocene event, shown with the yellow point. Very severe event. We no longer have this sampling spike in the Pleistocene. So this is the point I just made about the robustness of the analysis.

So what about extinction? Diversity curves aren't extinction curves. Fortunately, we can get an extinction curve out of this, because age ranges define extinction events. So here's the extinction rate curve for the Cenozoic, the last 65 million years. And it looks like there's a lot of up and down in this curve -- and it's easy to note that the points are higher in the Paleocene, in the very beginning. And, in fact, some of this up and down does mean something, and I've been able to determine that with a simple simulation test, which involves assuming a constant probability of extinction. And then, by randomization, simulating the number of extinctions you'd observe given a small and fixed number of species that are around at a particular time.

And what you see if you applied that test is that you need to throw out all of these points here, and a point here at 35 million years, and these two points in the end of the Miocene and the end-Pleistocene event, before you can fit the remaining points to a constant extinction probability model. So what that means is that there are a few extinction pulses in the past, and there are a lot of them in the Paleocene -- but that doesn't necessarily mean that the end-Pleistocene event is natural. And there are a couple of reasons. One of them is that the Paleocene points probably reflect the intrinsic turnover rate of the fauna, which is taxonomically very different from the modern fauna, and it probably does not represent a disturbance regime. The second reason that this occurrence of several extinction pulses doesn't tell us that the end-Pleistocene is natural, is that these are one-million-year sampling events. We're putting together all of the extinction events that occurred over a whole million years and calling that a particular rate. In the end-Pleistocene, however, we're talking about an event that happened at the scale of about 1,000 years, three orders of magnitude faster. We've even taken the preceding million years and made them a separate point. If we'd added them together, the end-Pleistocene would have been much higher.

So, essentially, by biasing the data as much as I could against finding the end-Pleistocene event to be unusual, I've still managed to find that the end-Pleistocene event was among the very worst in the post-Paleocene in the interval of normal turnover rates.

Now, there is a more interesting feature of the end-Pleistocene event -- namely, that it's selective against large animals. And the question that arises is: Is this a normal feature of extinction events? Do the large animals go out first when extinction gets really bad? So let's look at small mammals first. Here's a horned rodent from the Miocene. The rodent diversity pattern is very interesting. I won't discuss all these details. The important point is that the end-Pleistocene event had very little effect on the rodents.

... Now, this argument has been made by Russ Graham and others in the past. This isn't a new hypothesis that I'm testing -- it's a hypothesis out in the literature for many years. Here, again, is to indicate the type of geographic coverage we have -- not perfect, but pretty good. A fair number of samples. That's for the Pleistocene.

Now, fortunately for all of us, Dr. Graham and the FAUNMAP research group have been very gracious in making their data available on the web -- you can download them with the click of a mouse -- and their data comprise the last four time intervals that I'm going to focus on. This is years ago, in thousands of years; here are the time slices. My data are early Middle Pleistocene here, and last interglacial, roughly defined. The number of lists is pretty good in most of those intervals. It's great in the Late Holocene; it's not so great in the Early Holocene; it's not so great in the last interglacial. The number of species is pretty much not that variable -- it's pretty constant.

What I've done with those faunal lists is to look at what are called "conjunction patterns," which are patterns of overlap of geographic distributions between species. So the idea here is that, if you have a line, then the two species are found together in at least one place. If there's no line, as from this vole up here to this shrew down here, then those two species never occur in the same place. In the Recent we have complete data on this, because we have complete range maps that we can look at. So we can use the Recent as a template for the past in terms of these patterns of association.

If we do that, for example, for the late Wisconsinan, we see some interesting things. Most of the overlaps in the Recent are reproduced in the past -- all these blue lines. But some of them aren't, such as this conjunction -- which is called a disharmony, or disharmonious association -- between the species of vole, from the north, and this eastern shrew. And there's another disharmonious conjunction between the same vole and a wood rat.

So the basic idea is, we take these data and we turn them into an index of disharmony. The more red lines, the higher the index. We know that this means something -- it isn't just counting angels on the head of a pin -- because there's a lot of disjunction and a lot of conjunction in the recent. These patterns are very informative. If we look at the late Wisconsinan, although there are a lot of disharmonious conjunctions -- there's 178 in the FAUNMAP data -- that number is very small compared to the total number of overlaps of species.

So, relatively speaking, disharmony is a very rare pattern. We see the same thing throughout the six time slices, going from oldest to youngest. That pattern does not change very much, even across the mass extinction event here. It persists into the Late Holocene. We also see that the same species are involved in disharmony in different times. Late Wisconsinan versus Late Holocene -- this is very, very recent data. This is 10,000 and 12,000 years ago. The same species have the same rates of disharmony. This is a yellow-bellied marmot up here. What that means is that disharmony is not a unique feature of glacials or a unique feature of deglaciations -- it's a constant feature of the whole Quaternary. And we can show that by doing the same analyses on all pairs of the intervals. We see the same types of correlations over and over again, so essentially we see the same patterns persisting through time. And what that means is that disharmony is not an unnatural feature of the last glacial -- it's a natural feature of the Quaternary. And, in other words, what might be truly remarkable is range contractions in the last 500 years. Those might be responsible for the patterns of disharmony, and, essentially, the disharmony issue might be a red herring and not really germane to the extinction event.

So, to summarize all of these points, the extinction rates curves show that the end-Pleistocene event was among the worst ever in the Cenozoic. No other event before the end-Pleistocene was strongly selective for any group, such as large herbivores. We know that patterns of disharmony are real. They're relatively rare, although absolutely common. They're very predictable from one time to another; they're very constant through time. And what that means is, that the survivors of the event were not reorganized spatially -- they just kept on trucking -- and that makes it very hard to understand how habitat and climate changes could have done very much to the victims, whose ranges were changed and were, in fact, reduced to zero.

bio