A Crash Course in Multi-Level Selection Theory: Part 1-The Groundwork Laid by Dawkins and Gould
Responding to Stephen Jay Gould’s criticisms of his then most infamous book, Richard Dawkins writes in a footnote to the 1989 edition of The Selfish Gene, “I find his reasoning wrong but interesting, which, incidentally, he has been kind enough to tell me, is how he usually finds mine” (275). Dawkins’s idea was that evolution is, at its core, competition between genes with success measured in continued existence. Genes are replicators. Evolution is therefore best thought of as the outcome of this competition between replicators to keep on replicating. Gould’s response was that natural selection can’t possibly act on genes because genes are always buried in bodies. Those replicators always come grouped with other replicators and have only indirect effects on the bodies they ultimately serve as blueprints for. Natural selection, as Gould suggests, can’t “see” genes; it can only see, and act on, individuals.
The image of individual genes, plotting the course of their own survival, bears little relationship to developmental genetics as we understand it. Dawkins will need another metaphor: genes caucusing, forming alliances, showing deference for a chance to join a pact, gauging probable environments. But when you amalgamate so many genes and tie them together in hierarchical chains of action mediated by environments, we call the resultant object a body. (91)
Dawkins’ rebuttal, in both later editions of The Selfish Gene and in The Extended Phenotype, is, essentially, Duh—of course genes come grouped together with other genes and only ever evolve in context. But the important point is that individuals never replicate themselves. Bodies don’t create copies of themselves. Genes, on the other hand, do just that. Bodies are therefore best thought of as vehicles for these replicators.
As a subtle hint of his preeminent critic’s unreason, Dawkins quotes himself in his response to Gould, citing a passage Gould must’ve missed, in which the genes making up an individual organism’s genome are compared to the members of a rowing team. Each contributes to the success or failure of the team, but it’s still the individual members that are important. Dawkins describes how the concept of an “Evolutionarily Stable Strategy,” can be applied to a matter
arising from the analogy of oarsmen in a boat (representing genes in a body) needing a good team spirit. Genes are selected, not as “good” in isolation, but as good at working against the background of the other genes in the gene pool. A good gene must be compatible with and complementary to, the other genes with whom it has to share a long succession of bodies. A gene for plant-grinding teeth is a good gene in the gene pool of a herbivorous species, but a bad gene in the gene pool of a carnivorous species. (84)
Gould, in other words, isn’t telling Dawkins anything he hasn’t already considered. But does that mean Gould’s point is moot? Or does the rowing team analogy actually support his reasoning? In any case, they both agree that the idea of a “good gene” is meaningless without context.
The selfish gene idea has gone on to become the linchpin of research in many subfields of evolutionary biology, its main appeal being the ease with which it lends itself to mathematical modeling. If you want to know what traits are the most likely to evolve, you create a simulation in which individuals with various traits compete. Run the simulation and the outcome allows you to determine the relative probability of a given trait evolving in the context of individuals with other traits. You can then compare the statistical outcomes derived from the simulation with experimental data on how the actual animals behave. This sort of analysis relies on the assumption that the traits in question are both discrete and can be selected for, and this reasoning usually rest on the further assumption that the traits are, beyond a certain threshold probability, the end-product of chemical processes set in motion by a particular gene or set of genes. In reality, everyone acknowledges that this one-to-one correspondence between gene and trait—or constellation of genes and trait—seldom occurs. All genes can do is make their associated traits more likely to develop in specific environments. But if the sample size is large enough, meaning that the population you’re modeling is large enough, and if the interactions go through enough iterations, the complicating nuances will cancel out in the final statistical averaging.
Gould’s longstanding objection to this line of research—as productive as he acknowledged it could be—was that processes, and even events, like large-scale natural catastrophes, that occur at higher levels of analysis can be just as or more important than the shuffling of gene frequencies at the lowest level. It’s hardly irrelevant that Dawkins and most of his fellow ethologists who rely on his theories primarily study insects—relatively simple-bodied species that produce huge populations and have rapid generational turnover. Gould, on the other hand, focused his research on the evolution of snail shells. And he kept his eye throughout his career on the big picture of how evolution worked over vast periods of time. As a paleontologist, he found himself looking at trends in the fossil record that didn’t seem to follow the expected patterns of continual, gradual development within species. In fact, the fossil records of most lineages seem to be characterized by long periods of slow or no change followed by sudden disruptions—a pattern he and Niles Eldredge refer to as punctuated equilibrium. In working out an explanation for this pattern, Eldredge and Gould did Dawkins one better: sure, genes are capable of a sort of immortality, they reasoned, but then so are species. Evolution then isn’t just driven by competition between genes or individuals; something like species selection must also be taking place.
Dawkins accepted this reasoning up to a point, seeing that it probably even goes some way toward explaining the patterns that often emerge in the fossil record. But whereas Gould believed there was so much randomness at play in large populations that small differences would tend to cancel out, and that “speciation events”—periods when displacement or catastrophe led to smaller group sizes—were necessary for variations to take hold in the population, Dawkins thought it unlikely that variations really do cancel each other out even in large groups. This is because he knows of several examples of “evolutionary arms races,” multigenerational exchanges in which a small change leads to a big advantage, which in turn leads to a ratcheting up of the trait in question as all the individuals in the population are now competing in a changed context. Sexual selection, based on competition for reproductive access to females, is a common cause of arms races. That’s why extreme traits in the form of plumage or body size or antlers are easy to point to. Once you allow for this type of change within populations, you are forced to conclude that gene-level selection is much more powerful and important than species-level selection. As Dawkins explains in The Extended Phenotype,
Accepting Eldredge and Gould’s belief that natural selection is a general theory that can be phrased on many levels, the putting together of a certain quantity of evolutionary change demands a certain minimum number of selective replicator-eliminations. Whether the replicators that are selectively eliminated are genes or species, a simple evolutionary change requires only a few replicator substitutions. A large number of replicator substitutions, however, are needed for the evolution of a complex adaptation. The minimum replacement cycle time when we consider the gene as replicator is one individual generation, from zygote to zygote. It is measured in years or months, or smaller time units. Even in the largest organisms it is measured in only tens of years. When we consider the species as replicator, on the other hand, the replacement cycle time is the interval from speciation event to speciation event, and may be measured in thousands of years, tens of thousands, hundreds of thousands. In any given period of geological time, the number of selective species extinctions that can have taken place is many orders of magnitude less than the number of selective allele replacements that can have taken place. (106)
This reasoning, however, applies only to features and traits that are under intense selection pressure. So in determining whether a given trait arose through a process of gene selection or species selection you would first have to know certain features about the nature of that trait: how much of an advantage it confers if any, how widely members of the population vary in terms of it, and what types of countervailing forces might cancel out or intensify the selection pressure.
The main difference between Dawkins’s and Gould’s approaches to evolutionary questions is that Dawkins prefers to frame answers in terms of the relative success of competing genes while Gould prefers to frame them in terms of historical outcomes. Dawkins would explain a wasp’s behavior by pointing out that behaving that way ensures copies of the wasp’s genes will persist in the population. Gould would explain the shape of some mammalian skull by pointing out how contingent that shape is on the skulls of earlier creatures in the lineage. Dawkins knows history is important. Gould knows gene competition is important. The difference is in the relative weights given to each. Dawkins might challenge Gould, “Gene selection explains self-sacrifice for the sake of close relatives, who carry many of the same genes”—an idea known as kin selection—“what does your historical approach say about that?” Gould might then point to the tiny forelimbs of a tyrannosaurus, or the original emergence of feathers (which were probably sported by some other dinosaur) and challenge Dawkins, “Account for that in terms of gene competition.”
The area where these different perspectives came into the most direct conflict was sociobiology, which later developed into evolutionary psychology. This is a field in which theorists steeped in selfish gene thinking look at human social behavior and see in it the end product of gene competition. Behaviors are treated as traits, traits are assumed to have a genetic basis, and, since the genes involved exist because they outcompeted other genes producing other traits, their continuing existence suggests that the traits are adaptive, i.e. that they somehow make the continued existence of the associated genes more likely. The task of the evolutionary psychologist is to work out how. This was in fact the approach ethologists had been applying, primarily to insects, for decades.
E.O. Wilson, a renowned specialist on ant behavior, was the first to apply it to humans in his book Sociobiology, and in a later book, On Human Nature, which won him the Pulitzer. But the assumption that human behavior is somehow fixed to genes and that it always serves to benefit those genes was anathema to Gould. If ever there were a creature for whom the causal chain from gene to trait or behavior was too long and complex for the standard ethological approaches to yield valid insights, it had to be humans.
Gould famously compared evolutionary psychological theories to the “Just-so” stories of Kipling, suggesting they relied on far too many shaky assumptions and made use of far too little evidence. From Gould’s perspective, any observable trait, in humans or any other species, was just as likely to have no effect on fitness at all as it was to be adaptive. For one thing, the trait could be a byproduct of some other trait that’s adaptive; it could have been selected for indirectly. Or it could emerge from essentially random fluctuations in gene frequencies that take hold in populations because they neither help nor hinder survival and reproduction. And in humans of course there are things like cultural traditions, forethought, and technological intervention (as when a gene for near-sightedness is rendered moot with contact lenses). The debate got personal and heated, but in the end evolutionary psychology survived Gould’s criticisms. Outsiders could even be forgiven for suspecting that Gould actually helped the field by highlighting some of its weaknesses. He, in fact, didn’t object in principle to the study of human behavior from the perspective of biological evolution; he just believed the earliest attempts were far too facile. Still, there are grudges being harbored to this day.
Another way to look at the debate between Dawkins and Gould, one which lies at the heart of the current debate over group selection, is that Dawkins favored reductionism while Gould preferred holism. Dawkins always wants to get down to the most basic unit. His “‘central theorem’ of the extended phenotype” is that “An animal’s behaviour tends to maximize the survival of genes ‘for’ that behaviour, whether or not those genes happen to be in the body of the particular animal performing it” (233). Reductionism, despite its bad name, is an extremely successful approach to arriving at explanations, and it has a central role in science. Gould’s holistic approach, while more inclusive, is harder to quantify and harder to model. But there are several analogues to natural selection that suggest ways in which higher-order processes might be important for changes at lower orders. Regular interactions between bodies—or even between groups or populations of bodies—may be crucial in accounting for changes in gene frequencies the same way software can impact the functioning of hardware or symbolic thoughts can determine patterns of neural connections.
The question becomes whether or not higher-level processes operate regularly enough that their effects can’t safely be assumed to average out over time. One pitfall of selfish gene thinking is that it lends itself to the conflation of definitions and explanations. Evolution can be defined as changes in gene frequencies. But assuming a priori that competition at the level of genes causes those changes means running the risk of overlooking measurable outcomes of processes at higher levels. The debate, then, isn’t over whether evolution occurs at the level of genes—it has to—but rather over what processes lead to the changes. It could be argued that Gould, in his magnum opus The Structure of Evolutionary Theory, which was finished shortly before his death, forced Dawkins into making just this mistake. Responding to the book in an essay in his own book A Devil’s Chaplain, Dawkins writes,
Gould saw natural selection as operating on many levels in the hierarchy of life. Indeed it may, after a fashion, but I believe that such selection can have evolutionary consequences only when the entities selected consist of “replicators.” A replicator is a unit of coded information, of high fidelity but occasionally mutable, with some causal power over its own fate. Genes are such entities… Biological natural selection, at whatever level we may see it, results in evolutionary effects only insofar as it gives rise to changes in gene frequencies in gene pools. Gould, however, saw genes only as “book-keepers,” passively tracking the changes going on at other levels. In my view, whatever else genes are, they must be more than book-keepers, otherwise natural selection cannot work. If a genetic change has no causal influence on bodies, or at least on something that natural selection can “see,” natural selection cannot favour or disfavour it. No evolutionary change will result. (221-222)
Thus we come full circle as Dawkins comes dangerously close to acknowledging Gould’s original point about the selfish gene idea. With the book-keeper metaphor, Gould wasn’t suggesting that genes are perfectly inert. Of course, they cause something—but they don’t cause natural selection. Genes build bodies and influence behaviors, but natural selection acts on bodies and behaviors. Genes are the passive book-keepers with regard to the effects of natural selection, even though they’re active agents with regard to bodies. Again, the question becomes, do the processes that happen at higher levels of analysis operate with enough regularity to produce measurable changes in gene frequencies that a strict gene-level analysis would miss or obscure? Yes, evolution is genetic change. But the task of evolutionary biologists is to understand how those changes come about.
Gould died in May of 2002, in the middle of a correspondence he had been carrying on with Dawkins regarding how best to deal with an emerging creationist propaganda campaign called intelligent design, a set of ideas they both agreed were contemptible nonsense. These men were in many ways the opposing generals of the so-called Darwin Wars in the 1990s, but, as exasperated as they clearly got with each other’s writing at times, they always seemed genuinely interested and amused with what the other had to say. In his essay on Gould’s final work, Dawkins writes,
The Structure of Evolutionary Theory is such a massively powerful last word, it will keep us all busy replying to it for years. What a brilliant way for a scholar to go. I shall miss him. (222)
[I’ve narrowed the scope of this post to make the ideas as manageable as possible. This account of the debate leaves out many important names and is by no means comprehensive. A good first step if you’re interested in Dawkins’s and Gould’s ideas is to read The Selfish Gene and Full House.]
Read Part 2:
And Part 3: