Artificial Species Selection

Derek Turner writes . . .

Earlier this year, Joyce and I had some conversation about species selection (here and here). More recently, Leonard pointed out that Buffon may have anticipated the idea of species selection (here). This week, I thought I'd return to the topic and explain why I think species selection is such an important idea.

Species Selection: What is it?

Species selection occurs when the following conditions are met:

(1) There is differential persistence, extinction, and speciation (or branching) of whole lineages, or species, in evolutionary history.

(2) That process is not completely random. Instead, there is something about the species that affects their probabilities of persisting, going extinct, or speciating. In other words, you have differential fitness at the level of whole species.

(3) Whatever it is about the species in question that makes for differential fitness gets transmitted via speciation. So if species A gives rise to species B via ordinary speciation processes, features of A that increase (or decrease) extinction risk, say, will get passed on to B.

In other words, you need heritable variation in fitness at the level of whole species. In addition to these three conditions, some theorists think we need more than this for bona fide species selection:

(4) The traits of species that make the difference to their probabilities of extinction, persistence, and speciation, must be emergent traits. It’s not enough if those traits are merely aggregates of the traits of individual organisms.

Condition (4) takes us rapidly into philosophical territory. What does ‘emergence’ mean? It’s easier to give an example of an aggregate (non-emergent) trait. Many theorists have thought that large body size increases extinction risk. So big-bodied species, such as your sauropod dinosaurs or your woolly mammoths, might have lower species-level fitness. The problem, though, is that average body size is merely an aggregate measure. When we say that woolly mammoths are big, we don’t mean that the species itself is big, but that its individual members have a large body size on average. So if all we required for species selection is conditions (1) through (3), then large body size could be a factor. But if we insist on condition (4), large body size won’t cut it, because the trait is not emergent in the right sort of way.

Does large body size increase extinction risk? Average body size of the species is merely an aggregate trait. Image courtesy of wikimedia commons.

Does large body size increase extinction risk? Average body size of the species is merely an aggregate trait. Image courtesy of wikimedia commons.

There are trade-offs here. If you take the broader view, insisting only on conditions (1) through (3), then species selection is somewhat easier to find in nature, but it may also turn out to be less interesting, theoretically. On the other hand, if you insist on condition (4), then it’s really interesting, theoretically—because of the strong anti-reductionist implications—but also much tougher to document in nature.

Species Selection: Why Should We Believe in It?

As an empirical matter, though, on either the broader or the narrower view, it can be very difficult to point to cases in evolutionary history where we really need to invoke species selection in order to explain some puzzling phenomenon. I’m not sure if there is a consensus view about this at all—everything about species selection is controversial—but it might be good to start with what I’ll call the Explanation of Last Resort View.

“Sure, species selection could happen in principle. But it’s hard to document in any clear way, and it seems like the usual population biological explanations, which invoke selection, drift, mutation, and migration, give us a lot of explanatory mileage. So as a rule, don’t invoke higher level mechanisms like species selection unless you absolutely have to.”

Perhaps many scientists would add that as a matter of fact, you hardly ever, if ever, really need to invoke species selection.

Those scientists who’ve made the most convincing cases for species selection so far, have turned to the fossil record. For example, David Jablonski (1987) has done some research showing that the geographic range size of marine invertebrates makes a difference to extinction risk.

Charles Darwin. Image courtesy of wikimedia commons.

Charles Darwin. Image courtesy of wikimedia commons.

But why not take a lesson here from Darwin, that virtuoso of argument construction? Famously, in the Origin, Darwin tried to soften up resistance to his theory of natural selection by starting out with artificial selection. His go to example was pigeons, since he thought we have pretty good evidence that domesticated pigeon breeds are all descended from a common ancestor. Darwin also knew that the breeds of pigeons—the pouters and tumblers and fantails—were so different from one another that a naïve naturalist might well classify them as different species. In this case, it’s entirely plausible that artificial selection is the source of those remarkable differences. Generations of breeders determined the reproductive fate of their pigeons. If you accept this story about pigeons, then Darwin’s claims about natural selection in wild populations seem rationally irresistible.

A jacobin pigeon. Image courtesy of wikimedia commons.

A jacobin pigeon. Image courtesy of wikimedia commons.

A pouter pigeon. Image courtesy of wikimedia commons.

A pouter pigeon. Image courtesy of wikimedia commons.

One important piece of Darwin’s argument was the observation that human breeders—the artificial selectors—often have little if any clear idea of what they are doing. Selection is sometimes “unconscious”:

Thus, a man who intends keeping pointers naturally tries to get as good dogs as he can, and afterwards breeds from his own best dogs, but he has no wish or expectation of permanently altering the breed. Nevertheless, I cannot doubt that this process, continued during centuries, would improve and modify any breed …

Breeders are not necessarily aiming at any particular target. They are just making their own (likely aesthetic, perhaps also utilitarian) judgments about which of their dogs are the best. It’s a small step from “unconscious” human selection to mindless natural selection.

Paleontology and Neontology

Now to lay my cards on the table: It seems to me that if one wanted to defend species selection, the obvious way to do it would be to start just as Darwin did, but with cases of artificial species selection. Are there cases in which human activities contribute to the differential survival, reproduction, and speciation of whole lineages? And note that following Darwin, there’s no need to suppose that the human selectors are fully cognizant of what they are doing. Darwin seems to use “artificial selection” in a very broad sense: although his examples all involve animal breeding, once you see what he says about unconscious selection, it’s hard to see why we should extend the term “artificial selection” to other cases. For example, human fishing has led to size decrease in many fish species. That’s not a case of domesticated animal breeding, but it still seems like artificial selection.

Does artificial species selection occur in nature? Over the last few centuries, human activities have contributed significantly to the differential persistence and extinction of lineages, much as Darwin’s pigeon fanciers contributed to the differential reproduction of their birds. Plant and animal species living on islands, for example, have fared very, very badly, as humans have arrived in place after place with cats, rats, mice, pigs, goats and other ecological troublemakers. Species on continental landmasses have tended to do better. Human activities have meant that island dwelling species have higher extinction risk. This sort of case clearly meets conditions (1) through (3) above, though perhaps not condition (4). The point, though, is that we can explain the extinction patterns by making high-level generalizations about the extinction risk associated with living on islands. And it's human activity that makes island-dwelling risky.

This notion of artificial species selection is not really new. In an important review paper on species selection research, David Jablonski (2008) wrote that “[t]oday’s biota appears to be in the midst of a massive experiment in strict-sense species selection” (p. 515). That seems right, although it would take more work to show that condition (4) is met in cases of artificial species selection. Species selection is real. It’s happening right now.

Why is this not totally obvious? I’ll conclude on a speculative note: In some of his recent contributions, Leonard has stressed some of the conceptual discontinuities between paleontology and neontology (here). Species selection theory is a product of the paleobiological revolution of the 1970s and 1980s. It was devised and defended by paleontologists (Steven Stanley, Stephen Jay Gould, Elisabeth Vrba, as well as the philosopher Elisabeth Lloyd), for the purpose of explaining the patterns of life’s history. The failure to recognize the obvious relevance of species selection theory to conservation biology seems like another instance of the all too familiar paleo/neo disconnect.


Darwin, C. (1859/1964), On The Origin of Species, A Facsimile of the First Edition. Harvard University Press.

Jablonski, D. (1987), "Heritability at the species level: analysis of geographic ranges of cretaceous mollusks," Science 238: 360-363.

Jablonski, D. (2008), "Species selection: Theory and data," Annual Review of Ecology and Systematics 39: 501-524.

The argument in this post is developed in greater detail here.


Extinction and Expiration Dates

A Conversation with Joyce Havstad about Species Selection

A coelacanth at the Naturhistorisches Museum in Vienna. Some species just seem to go on and on . . . and on and on. They don't have expiration dates. Is this a problem for the theory of species selection?

A coelacanth at the Naturhistorisches Museum in Vienna. Some species just seem to go on and on . . . and on and on. They don't have expiration dates. Is this a problem for the theory of species selection?

Derek Turner writes ...

Joyce Havstad offers an incredibly careful and generous discussion of my Paleontology book. (Hi Joyce – and thanks!) Every philosopher should be lucky enough to get such careful and challenging feedback. Joyce makes some critical points and raises some questions about important technical details. Here I want to pick up on just one of the challenges that she raises. In the book I am pretty sympathetic toward the theory of species selection. Joyce, however, highlights a really interesting problem for that theory, one that I don’t consider in the book at all. Nor do I know of anyone else who discusses it. (Readers, if you know of any scientists or philosophers who have thought about this, can you please share?) It seems like this problem, which I’ll call the Expiration Date Problem, needs some attention.


What is Species Selection all About?

When Darwin formulated the theory of natural selection, he was thinking about the differential survival and reproduction of individuals within a population. But what if whole species are going through a similar process at a higher level? Differential speciation and extinction seem a lot like differential reproduction and survival.

Steven Stanley, a paleontologist, described species selection in a paper he published in 1975:

In this higher-level process species become analogous to individuals, and speciation replaces reproduction. The random aspects of speciation take the place of mutation. Whereas, natural selection operates upon individuals within populations, species selection operates upon species within higher taxa, determining statistical trends. In natural selection types of individuals are favored that tend to (A) survive to reproduction age and (B) exhibit high fecundity. The two comparable traits of species selection are (A) survival for long periods, which increases the chance of speciation, and (B) the tendency to speciate at high rates. Extinction, of course, replaces death in the analogy (p. 648).[1]

Notice how Stanley draws a parallel between the following two phenomena:

(A)  An organism surviving to reproductive age.

(A*) A whole species surviving for a long period, “which increases the chance of speciation”

Joyce’s worry—the problem of expiration dates—is that there is a relevant difference between organisms and species that may cause trouble for Stanley’s analogy here.


The Problem of Expiration Dates

Joyce writes:

By far the majority of organisms have an unyielding expiration date.  There’s an upper bound on how long they can live before they die.  Organismal fitness is therefore constituted by both survival and reproduction, but the former is mostly important as a way of guaranteeing the latter.  Species, however, can perdure—and therefore, survival can be a more independent and significant contributor to fitness.”

This is a really good point. Ordinary organisms have a maximum lifespan. But species can, at least in principle, last indefinitely. Species have no “unyielding expiration date,” as Joyce aptly puts it. Perhaps some “living fossil” species are good examples of this—think of horseshoe crabs or the coelacanth pictured above. Some species do seem to persist for many millions of years, with no clear upper bound on how long they may last. The question is whether this difference matters, and if so, how.

Joyce argues that this difference between species and individual organisms might affect how we think about some of the famous alleged cases of species selection, such as Elisabeth Vrba’s case of the African antelopes.

Vrba compared impalas with wildebeests over the last several million years.[2] Although impalas are more abundant (in the sense that there are more individual animals), wildebeests are a more species-rich group. In the last five or six million years, impalas have barely speciated at all, whereas there are many more species of wildebeests. This looks like an interesting case where species-level fitness and organism-level fitness pull apart. And on more liberal conceptions of species selection, that pulling apart is what you really need. Their abundance suggests that individual impalas are doing pretty well in their environment, but the low speciation rate suggests a lower species-level fitness. 

(One qualification: Vrba herself did not really think of this case as a case of species selection. She called it “effect macroevolution,” because she thought that the differential speciation rates of impalas vs. wildebeests were just a side effect of ordinary microevolutionary processes. But some people who take a more liberal view of what counts as species selection might consider this to be a good example of it.)

Joyce raises an important conceptual question: Is it really correct to say that the impalas have a lower species-level fitness? We need to be careful to avoid thinking of species-level fitness as nothing more than speciation propensity. Persistence (the species-level analogue of survival) matters too! And this is where the disanalogy mentioned above starts to become an issue. If the impala species had an upper bound on how long they could last—if they were really like individual organisms—then the low speciation rate really would make for lower species-level fitness. But of course with whole species, there’s no upper bound. The impalas might have a much lower extinction risk than the wildebeest species do. In fact this is really plausible. Part of Vrba’s original argument was that the impalas are ecological generalists, while the wildebeest species tend to specialize more on particular food sources. In general, generalists have lower extinction risk. So taking Joyce’s point into account, it could be that Vrba’s (and my) initial take on this case is not quite right: Maybe the impalas are not less fit at the species level than the wildebeests at all!


Why do we need the theory of species selection?

Species-level fitness has two ingredients: speciation propensity and extinction risk. How do those fit together, in Vrba’s case, and others? The lack of expiration dates makes this question tough to answer.

Now here is a philosophical trial balloon: One possible response to this problem might be to stop worrying about estimating species level fitness. Instead, perhaps, the right approach is to try to estimate speciation propensity and extinction risk independently, treating those as two different factors that can make a difference to macroevolutionary patterns. Perhaps one could concede the difficulty of saying exactly how to combine these into a single quantity (species-level fitness), while arguing that treating them independently still gets us most of what we could reasonably want from the theory of species selection.

This response requires us to say a bit about what the theory of species selection is for. Here I think the theory might have payoff in two different domains:

(1)   Species selection gives us one possible way of explaining certain large-scale patterns and trends in evolutionary history.

(2)   Species selection also gives us a useful (possibly predictive) perspective on the current biodiversity crisis, because we are now in a period of artificial species selection, where human activities are biasing macroevolutionary processes.

The second point is especially important, though it remains under-explored. In an important 2008 review paper, David Jablonski wrote that “[t]oday’s biota appears to be in the midst of a massive experiment in strict-sense species selection” (p. 515).[3] If Jablonski is right, species selection theory could be another paleontological contribution to conservation biology.

It might be possible to make good on both projects (1) and (2) without really solving the problem of expiration dates. For example, it might be possible to generate interesting explanations of large-scale evolutionary patterns merely by estimating differential extinction risk. If burrowing animals have a lower extinction risk when a meteoroid hits, that alone can help explain resulting patterns in the fossil record. And for that explanatory purpose, it might not matter much how extinction risk combines with speciation propensity to constitute species-level fitness. Similarly, with respect to project (2), it could be really useful to estimate extinction risks of different species, even without worrying so much about how the extinction risk goes together with speciation propensity to constitute species-level fitness.

In other words—and this is just a trial balloon—maybe the ingredients of species-level fitness are more interesting and important than speciesl-level fitness itself.


[1] Stanley, S. (1975), “A Theory of Evolution Above the Species Level,” Proceedings of the National Academy of Sciences 72(2): 6467-650.

[2] Vrba, E. (1987), “Ecology in relation to speciation rates: Some case histories of Miocene-Recent mammal clades,” Evolutionary Ecology 1: 283-300.

[3] Jablonski, D. (2008), “Species Selection: Theory and Data,” Annual Review of Ecology, Evolution, and Systematics 39: 501-524.