The short answer: through kin selection.

According to the new paper from the Proceedings of the National Academy of Sciences (PNAS) shown below, and in general in evolutionary biology, altruism is defined as “a behavior decreasing the expected survival and/or reproduction (fitness) of the actor while increasing the fitness of the recipient.”

The simplest example of such altruism involves parental care. A human mother taking care of her child is using resources (milk, time, effort) that in fact reduces her chance of survival or of having future kids. But the kid itself, the recipient, benefits. Parental care evolves because the cost to the mom is less than the benefit to the kid she tends.

Likewise with any sacrifice people make for their relatives. The reason this has evolved is that genes promoting parental behavior do entail a cost to their carriers, but they more than repay that cost by helping the perpetuation of the same genes (“genes identical by descent”) in the offspring, which has a 50% of getting a parental-care gene from the parent. Thus the gene gets a net boost from the behavior it produces.

So there’s a calculus involved for genes that reduce your fitness but help that of the recipient. This calculus is expressed in “Hamilton’s rule,” introduced by the great evolutionary biologist W. D. Hamilton. In general, a gene producing altruistic behavior—reducing the fitness of its carrier but helping others who carry copies of the same gene—will evolve by natural selection (i.e., increase in frequency) if it satisfies this equation:

r  x b > c

where c is the fitness cost to the donor of performing the act, b is the benefit to the recipient, and r is the “degree of relationship”, i.e., the chance that the recipient actually carries a copy of the altruism-producing gene because it’s related to the donor (“identity by descent”).

So, for example, r for parents vs. offspring is 0.5: the chance that an offspring will inherit an altruism gene (gene form, actually: an “allele”) from a parent is 50% due to segregation and assortment during reproduction. One can conclude that a gene that makes you expend effort to help your kid will be favored by natural selection if the fitness benefit to your kid is at least twice the cost to you. r for siblings is also 50% (brothers and sisters share half their genes), so a gene could be favored that causes you to help your siblings if the cost to you is also less than half the benefit to your siblings. r for uncles compared to nieces and nephews is 25% (therefore, for Uncle Joe, his altruism will evolve if the cost to him is less than a quarter of the benefit to niece Sarah, and so on.

The interaction between relatives, close or distant gentically, is the way that most evolutionists think that altruism has evolved. For a gene that incurs fitness costs in its bearer, but doesn’t give a benefit to those carrying other copies of the same gene, will go extinct. This is why when we observe self-sacrifice in nature, it’s nearly always to help relatives. (Think of the “broken wing” display in which a mother bird, feigning injury but risking her life, lures a predator away from her chicks.)

And when animals have a way to recognize and avoid taking care of unrelated organisms, they can. Here’s a note evolutionist Bruce Lyon sent me about the work of him and his colleagues on coots:

American coot females lay eggs in each others’ nests and they recognize and the host parents deal with the brood parasitic eggs/offspring at two stages: they recognized about a third of parasitic eggs and reject them by burying them down in the nest and they can also learn to recognize some parasitic chicks, and if they recognize the chicks they kill them.

. . . Lots of other birds have been shown to be able to recognize their own chicks, as in colonial seabirds, but they don’t use this to kill other chicks but instead insure that they feed their own kids.

This makes no sense unless parental care involves relatedness.  (If you have questions about this, I’ll ask Bruce to answer them in the comments.)

It doesn’t have to be direct relatedness, either. If a population is viscous, with individuals not moving around much, people will become related simply because they mate more often with nearby individuals. That’s why there’s a high degree of relatedness in small religious communities like the Dunkers and Amish, who don’t marry their siblings or cousins but marry those in the community. Over time, this causes an increase in relatedness in such communities.

Hamilton proposed his “rule” in 1964, but others hit on it as well, including J. B. S. Haldane, who was reported to say that he’d lay down his life for two brothers or eight cousins (you’d have to save all the relatives’ lives for this to work), and the idea was also worked out mathematically by the eccentric biologist George Price.

But in the last two decades, several biologists have claimed that altruism could evolve without this kind of kin selection—without individuals behaving in a way to favor their relatives. Most prominent among these contrarian biologists is Martin Nowak at Harvard, who has said that altruism doesn’t need relatedness to evolve, simply requiring a particular population structure. Other biologists have said that kin selection could work, but so could population structure alone.

It turns out that in all these cases, the population structure proposed in fact causes individuals to be related and favors altruism because of that relatedness. While many biologists recognize the mathematical equivalence of “population structure” and “kin selection models”, Nowak has denied this, stating that geographic population structure alone (his model of “spatial selection”), even if it doesn’t create a web of relatedness, could favor the evolution of altruism.

Nowak is wrong. This is demonstrated in the new paper in PNAS by Kay, Keller, and Lehmann (click on link to get it, pdf here, and reference at bottom). The upshot: you can’t get the evolution of altruism with population-structure alone, unless that population structure creates kin relationships that satisfy Hamilton’s rule. Kin selection remains the sine qua non for the evolution of altruism.

 

What the authors did is simple: they looked up all the scientific papers that showed the evolution of altruism, including those that ignored kin selection as well as those that denied kin selection was operating, and then analyzed whether the models indeed created a structure in which relatedness was important to determine whether kin selection—even if ignored or denied—was crucial for evolving altruism. They found 89 papers of theoretical models in which altruism evolved. The authors parsed them this way (my emphasis):

Among the 89 altruism models, 46 adopted Hamilton’s conceptual framework, attributing the evolution of altruism to positive relatedness. The remaining 43 all claimed alternative mechanisms. To evaluate the veracity of their claims, we first subdivided these 43 papers into those where the role of relatedness was denied (17 cases; SI Appendix, Table S3), and those which made little or no mention of relatedness (26 cases; SI Appendix, Table S4).

Among the 17 papers where the presence/role of relatedness was denied (SI Appendix, Table S3), our analysis of the life cycles of the models showed that the proposed scenario led to positive relatedness between interacting agents in every case. Moreover, in most of these models, agents reproduced clonally (e.g., “parents pass on their type to their offspring”) with interactions occurring among nearest neighbors, as in the stepping-stone model of Fig. 1, with only one individual per node/group. This represents the tritest instance of kin selection.

As for the remaining 26 models which proposed non-kin mechanisms for the evolution of altruism but didn’t mention relatedness, this is what Kay et al.’s analysis showed:

The 26 papers which make little or no mention of relatedness attribute the evolution of altruism to diverse alternative mechanisms including “social diversity,” “social viscosity,” “topological heterogeneity,” “network heterogeneity,” “network reciprocity,” “spatial reciprocity,” “spatial structure,” and “multiplex structure” (SI Appendix, Table S4). Analysis of these models revealed that in every case interacting individuals are related, relatives benefit from each other’s altruism, and kin selection therefore operates.

So none—zero, zip, bukes—of the “alternative” models could evolve altruism without kin selection and relatedness. This isn’t so mendacious when the authors just ignore kin selection, proposing models that nevertheless produce the interactions that allow kin selection. But it IS bad behavior when authors like Nowak claim that the evolution of altruism has nothing to do with kin selection and relatedness. That is a form of careerism—proposing some new mechanism when you haven’t done the scientific legwork (like Kay et al. did) to show that the new boss is the same as the old boss.

The upshot: the evolution of biological altruism, in which individuals sacrifice their own fitness to help others, cannot proceed without kin selection. There would be no selection on parents to help adopted children, since they aren’t related. (The fact that they do, in both humans and animals, is certainly a case of misplaced parental instinct. Warblers feeding cuckoo chicks, who aren’t even in the same species, is a prime example of hijacking of parental impulses.)

Why do so many authors ignore kin selection or say it isn’t operating in the evolution of altruism? (Nowak isn’t the only one of the latter.) Kay et al. give three suggestions. The first is careerism, as I’ve mentioned above: you don’t get famous by just showing what’s already been demonstrated. But they also note that some models are made by people who aren’t evolutionists and thus may be unaware of Hamilton and Price’s work (these people are economists, physicists, and so on). Finally, some authors know and understand Hamilton’s rule but are so steeped in it that they simply don’t bother to bring it up explicitly in their models.

So don’t believe claims that altruism can evolve without kin selection.

BUT, what about those cases in which, say, humans help others, risking their lives for people who aren’t related to them? I often use as examples volunteer firemen, who risk their lives for people they don’t even know. Or, in war, soldiers have died by throwing themselves on a grenade to save their platoon. This is certainly altruism, but it doesn’t involve kin. This kind of sacrifice is almost completely unknown in other species, where individuals aren’t seen to risk their lives for non-relatives. That alone gives you a clue that there is some cultural aspect to this kind of altruism in humans. But you are as good as I am in speculating about this, and I’ll leave it as a thought exercise.

Coda: I never met Hamilton even though we overlapped in time (he lived from 1936-2000, dying at only 63 from what may have been a combination of an ulcer and malaria). But I know many people who knew Hamilton well, and without exception they paint him as an unassuming and genial man—an all-around nice guy as well as a scientific genius (he was also a keen naturalist and spent a lot of time in the tropics). He had some bizarre ideas, but also some ideas that became foundational in the evolution of behavior. Here he is (I just noticed that he looks a bit like me, but with longer hair).

________________

Kay, T., L. Keller, and L. Lehmann. 2020. The evolution of altruism and the serial rediscovery of the role of relatedness. Proceedings of the National Academy of Sciences 117:28894-28898.

In 2010, Martin Nowak, Corina Tarnita, and E. O. Wilson wrote a paper in Nature (reference and link below) arguing that “kin selection,” selection based on relatedness (shared alleles among nestmates) was not—as had long been maintained—a key factor in the evolution of “eusocial” insects. (Those are species in which there are nonreproductive “castes” of workers, with some tending the brood) while reproduction is limited to one or a few “queens”.)

The problem with this paper was their dismissal of relatedness as an important factor in the evolution of this remarkable social system (eusociality isn’t just limited to insects; we also see it in some crustaceans and in naked mole rats). Nowak et al.’s “model,” such as it was, did not allow the degree of relatedness to vary, so there was simply no basis for their claim that relatedness was not “causal” in the evolution of eusociality. In fact, there was already evidence that kin selection was important in the evolution of eusociality. As I wrote in March of this year:

But the evidence for kin selection and relatedness is still clear. For example, eusociality in Hymenoptera has evolved several times, but always occurred in an ancestral lineage in which queens mated singly rather than multiply: a statistically significant finding (Hughes 2008; reference and free link below). That’s important because in such cases offspring are more related to each other than are offspring produced by different fathers. Further, Bob Trivers showed that other patterns in bees, ants, and wasps—especially the observed ratios of males to reproductive females in colonies—also followed the dictates of what kin selection predicted. There are still other behavioral recognition experiments of kin versus non-kin supporting the importance of relatedness.

The 2010 paper by Nowak et al. was criticized on these and other grounds by virtually every evolutionary biologist working on the evolution of social behavior. One critique had over 130 authors! But Nowak et al. have stood their ground, largely alone in their views with the exception of David Sloan Wilson, who, for reasons I can’t fathom, argues that Nowak et al.’s “group selection” argument is right, and has just published a note on his website called “Mopping up final opposition to group selection.”  As the battle winds down, D. S. Wilson has declared victory for the wrong side!

I’ve continued to monitor the controversy, and you can find the links to my many posts here. The latest critique of Nowak et al. was leveled by Liao, Rong, and Queller (link and reference below), which I reported on here.  I won’t go into their findings in detail, but Liao et al. did what Nowak et al. should have: they made models in which relatedness was allowed to vary, for that’s the only way to see how important kin selection (i.e., selection based on relatedness) is in the evolution of eusociality. As I wrote in that post:

Unlike Nowak et al., Liao et al. varied total relatedness by allowing a certain fraction of offspring in the nest to be unrelated to the “queen” rather than simply her clones. (This could occur by immigration of insects from other nests, or by queens laying eggs in other queens’ nests.) What they found is that relatedness indeed makes a big difference: under conditions in which worker behavior is affected by their own genes rather than just the queen’s, eusociality evolves much more easily when relatedness between queen and “worker” is higher. In other words, higher relatedness (kin selection) is causal in this circumstance, not just a consequence of the evolution of eusociality. Nowak et al. were wrong, and all the statements of this group about the uselessness of kin selection based on this model are also wrong.

Nevertheless, Nowak and his colleagues are showing a characteristic trait of some scientists: a complete refusal to admit that their critics had any valid points at all. In response to the Queller et al. paper, Nowak and Benjamin Allen just published a note in PLoS Biology (reference and link below) defending the original result and dismissing Liao et al.’s criticisms.  Responding to that in a one-page note in the same issue, Queller, Rong, and Liao once again show how Nowak et al. (2010) were misguided and misleading, and that the subsequent Nowak and Allen paper apparently concedes ground while pretending not to do so.

Here is what Nowak and Allen now contend (I’m summarizing what I see as the two important points):

1. Nowak et al. never said that relatedness was unimportant. From their paper (“LRQ” is Liao, Rong, and Queller’s paper modeling variation in relatedness; “NTW” is Nowak, Tarnita, and Wilson’s original paper):

Why do LRQ investigate such models? They present NTW as saying relatedness does not matter in general, but this is incorrect. Instead NTW write, “Relatedness does not drive the evolution of eusociality. We can use our model to study the fate of eusocial alleles that arise in thousands of different presocial species with haplodiploid genetics and progressive provisioning. In some of those species eusociality might evolve, while in others it does not. Whether or not eusociality evolves depends on the demographic parameters of the queen (…), but not on relatedness. The relatedness parameters would be the same for all species under consideration”

Nowak and Allen also note (see their Figure 1), that among the many species that have the kind of mother/offspring association that could promote eusociality (“progressive provisioning,” in which offspring are continuously fed in nests), only a few have evolved eusociality.

I see this as disingenuous. NTW did indeed argue that relatedness is unimportant in the evolution of eusociality, precisely the problem that LRQ investigated, showing that relatedness was important. As for the fact that eusociality didn’t evolve in a lot of progressive-provisioning species, everyone, including Queller and his colleagues (and me!) admits that factors other than relatedness can influence the evolution of eusociality. After all, there are ecological factors that affect the benefits and the costs of evolving sterile castes, fertile queens, and the like. But the results of Hughes et al. and of Trivers suggest strongly that kin selection was important in the evolution of eusociality. Neither NTW nor Nowak and Allen mention these results. Leaving out discussion of results that support your opponents’ position is not a good way to behave in science.

2. Nowak and Allen argue that Liao et al.’s models of varying relatedness are biologically unrealistic. You can read their criticisms themselves, and I’m unable to judge, not knowing much about the biology of the Hymenoptera, whether these particular models correspond to situations that actually obtain in nature.

I asked my friend Phil Ward, a professor of entomology at the University of California who works on Hymenoptera, about this issue, and he replied that while Liao’s “mixing model” seems a bit contrived, “there is probably enough nest usurpation and nest-sharing among non-social bees and wasps to generate significant variation in relatedness among interacting groups of individuals, if not exactly in the manner modeled by Liao et al. (2015).”

I agree. Surely the degree of relatedness can vary among nests in nature, however that happens, and if relatedness is “causal,” (which Nowak et al. deny but Liao et al. affirm), then that will affect the likelihood of evolving eusociality. To dissect the specific models without addressing whether something might alter relatedness in non-social hymenopteran nests is to throw out the baby with the bathwater.

That said, we clearly need more empirical work on the biology of non-social Hymenoptera that build nests so that we can answer the question that Nowak and Allen (and many others) have posed: Why have most of these species not evolved eusociality? The answer likely involves some combination of ecology, behavior, and relatedness.

In their very short response to Nowak and Allen, Queller et al. can be quoted directly, as their points are clear:

We asked whether the model of Nowak, Tarnita, and Wilson (NTW), when applied to their chosen test case of eusociality, makes any important difference. Does it refute kin selection theory? Does it offer new insights? The answer to both questions is no.

I agree with that statement.  They go on (my emphasis):

Now Nowak and Allen suggest that we have misinterpreted NTW. For example, NTW did not mean that relatedness is unimportant. Instead, they only meant that if relatedness is high and held constant, other factors determine which species evolve eusociality, and that this is an issue the kin selectionists have not considered. On the contrary, it is completely obvious from Hamilton’s rule; if you hold relatedness constant, differences will be determined by variation in costs and benefits. There have also been more specific studies about synergistic factors affecting these costs and benefits. Moreover, if this is the basis for NTW’s claim that relatedness is not causal, then we have shown that NTW’s other parameters are also not causal, because when we force them to be constant, only variation in relatedness matters. Finally, this apparent concession about the importance of relatedness is perplexing, given that Nowak and Allen expend significant effort questioning the details of exactly how we modeled lower relatedness, while continuing to equivocate about the real issue of how relatedness matters. Low relatedness groups are real and can be formed in many ways, but with offspring control they do not give rise to eusociality. If Nowak and Allen think otherwise and believe that there are reasonable ways to lower relatedness so that it does not make eusociality harder to evolve, then they should show how.

This is telling. NTW truly equivocate about the notion of “causality”, using a double standard when assessing relatedness versus ecological factors. In fact, as Queller notes, both ecology and relatedness can be “causal” in the sense that, if other things are held equal, variation in these factors can both tip the balance toward the evolution of eusociality. The question is whether relatedness did tip the balance, and the results of Hughes et al. (2008) suggest that it did.

Finally, Queller et al. end their response like this:

. . . If NTW did not actually mean that relatedness is unimportant, and if they did not mean that workers are merely robotic extra-somatic projections of the queen’s genome, and if they did not mean that eusociality was as hard to evolve as suggested in their main examples, then we are in happy agreement! But if this is so, why do they not just explicitly say, for example, “our method agrees with inclusive fitness in showing that higher relatedness is crucial in the evolution of eusociality”? Perhaps because it would require admitting that what we have learned about eusociality from kin selection models still stands, and that the NTW models, despite their much greater complexity, have so far added little more.

This is as close as Queller, a soft-spoken guy who doesn’t like controversy, can come to calling his opponents misguided but ambitious scientists who won’t admit that they’ve distorted the situation. That, at least, is my take on the exchange. Nowak and D. S. Wilson have staked their careers on the “kin-selection-is-wrong-and-my-theory-is-better” view, and they’re obdurate about that. But such stubbornness is more akin to theology than to science.

 ______

Hughes, W. O. H., B. P. Oldroyd, M. Beekman, and F. L. W. Ratnieks. 2008. Ancestral monogamy shows kin selection is key to the evolution of eusociality. Science 320:1213-1216.

Liao, X., Rong, S., and D. Queller, 2015. Relatedness, conflict, and the evolution of eusociality. PLOS Biology | DOI:10.1371/

Nowak, M. A., C. E. Tarnita and E. O. Wilson.  2010.  The evolution of eusociality.  Nature 466: 1057-1062.

Nowak MA and B. Allen (2015) Inclusive fitness theorizing invokes phenomena that are not relevant for the evolution of eusociality. PLoS Biol 13(4): e1002134. doi:10.1371/journal.pbio.1002134

Queller, D. C., S. Rong, and X. Liao. 2015. Some agreement on kin selection and eusociality? PLoS Biol 13(4): e1002133. doi: 10.1371/journal.pbio.1002133