Steven Pinker on “The False Allure of Group Selection”



Join 36.9K other subscribers

A recurrent idea in Steven Pinker’s essay is that group selection “adds little to what we have always called ‘history’.” I argue, on the contrary, that cultural multi-level selection (CMLS) provides a highly productive theoretical framework for the study of human history, including (and integrating over) both modeling and empirical approaches. Notable examples developed during the last decade include the evolution of religion (e.g., in the work of David Sloan Wilson and Richard Bellah), the evolution of monogamy (e.g., the work of Joseph Henrich), and my own work on the evolution of social complexity. It is worth noting that I came to CMLS indirectly. My own research has focused on the mechanisms that underlie the rise – and recurrent demise – of historical large-scale societies (‘megaempires’), and so far I found no other framework that could even approach the utility of CMLS.

Large-scale human societies are not simply undifferentiated ensembles of individuals. They are ‘segmentary’ in the anthropological jargon, that is, their internal structure can be represented as groups nested within groups nested within groups … and so on. In other words, human societies are truly multilevel entities and evolution of large-scale sociality in humans was not just one major evolutionary transition, but a whole cascade of them. In order for societies to exist without fragmenting, forces holding together groups at various levels must overcome centrifugal tendencies (and when they fail to do so, the result is a failed state). A major theoretical result in MLS is the Price equation, which specifies the conditions under which the balance shifts either toward integration, or disintegration.

Cultural traits of central interest are prosocial norms and institutions (for more on this, see recent articles by Peter Richerson, Sam Bowles, and others). They are critical for the stability and functioning of large-scale societies, but have very significant costs at lower levels of social organization. Thus, we have a typical multilevel situation, in which traits are under divergent selective pressures acting at different levels of organization.

CMLS is much more than a metaphor, it yields quantitative predictions that can be (and have been) tested with historical data. The Price equation includes not only coefficients of selective pressures (working against each other at lower vs. higher levels), but also cultural variances at two (or more) levels. Incidentally, the critical importance of variances is a new insight for most social scientists, not steeped in evolutionary theory. But we can go beyond such conceptual insights to empirical applications. In particular, the Price equation suggests that large states should arise in regions where culturally very different people are in contact, and where interpolity competition – warfare – is particularly intense.

In a recent article, I tested this prediction for the period of human history between 500 BCE and 1500 CE, and found a very good match between predictions and data. A further development of this approach is the current collaborative project with Tom Currie, Edward Turner, and Sergey Gavrilets. We have developed an agent-based model of cultural evolution of prosocial institutions on a realistic landscape (Afroeurasia between 1500 BCE and 1500 CE). We also quantified the spatiotemporal distribution of historical large-scale societies by counting how frequently each 100 × 100 km square found itself within a large territorial polity. Our results indicate that the model predicts over 60 percent of variance in the data, a level of precision practically unheard of in historical applications.

Thus, the theoretical framework of CMLS provides not only new conceptual insights into the study of human history, but it also guides empirical research and, most notably, yields predictions that exhibit an excellent match with data.

A note:

This is a draft of a commentary that I intend to submit to The Edge, where Steven Pinker’s essay was published. I would really appreciate any comments. The commentary is a bit too long (600 instead of 500 words), so suggestions on which pieces are not critical would also be very welcome.

An additional note:

After posting the commentary I discovered the “Editorial Marching Orders” for publishing comments at the Edge, which specifically prohibits ‘self-promotion’. As my counter-example to Pinker’s claim that group selection adds little to the study of human history relies on my own work, this disqualifies my comment for the Edge. So I will have to either rethink my commentary, or simply not bother submitting it to the Edge. In any case, I remain very interested in comments from all.

A (hopefully) final note:

I have now heard that the restriction against referring to own works has been relaxed, so from this point of view, my commentary should be OK after additional editing.

Notify of
Most Voted
Newest Oldest
Inline Feedbacks
View all comments
Yasha Hartberg

I think this is an excellent reply, Peter, and very interesting. It’s already quite concise, so it’s hard to know what to edit. Perhaps you could leave out the paragraph beginning, “Cultural traits of central interest?” That wouldn’t get you under 500 words, but it would get you closer to your goal.

Peter Turchin

Yasha, thanks. It looks like the word limit has been lifted, so current length is OK. In fact, I am thinking of adding a bit more on Pinker’s ‘grand deception’ theory, and on the evolution of monogamy (unless Joe Henrich decides to do it himself).

David Sloan Wilson

I also like this reply, Peter, and I’m glad that you’re addressing what Steve said about history. For me, the statement that “group selection adds little to what we have always called history” is like saying “evolution adds little to what we have always called the fossil record”. Both recorded human history and the biological fossil record are bodies of information that require theoretical frameworks to be understood. The study of human history lacks of theoretical framework, and evolutionary theory provides one.

Steve’s comments about history go beyond group selection. I think he has a hard time thinking about cultural evolution in general, multilevel or otherwise. He thinks that if psychological processes are involved, such as intentional decision-making, it no longer qualifies as evolution.

Peter Turchin

David: It looks like several of us will be taking Steven to task for his unreasonable requirement that variation has to be generated purely randomly. It’s not in the basic Darwinian postulates!

Mark Sloan

Peter, I have followed your blog for a while, and thought I would try posting.

I like your response to Pinker, but I had a different perspective. It seems to me that his focus on “natural selection” only behaviors creates a straw horse version of multi-level selection to attack.

For example, he ground rules out any counterexamples that are not based on “natural selection”. (I understand him to mean by “natural selection” something like selection without intent, or random selection, based only on reproductive fitness.)

This ground rule removes from consideration some of multi-level selections best supporting data: results of group selection experiments, which are almost inevitably based on the intent of the investigator, and enforced cultural norms (moral standards) within groups including nation states. Pinker ground rules out cultural norms because they are both chosen based on people’s intent, and are selected for by many benefits of increased cooperation, not necessarily reproductive fitness.

Also, the expansion of the definition of kin-selection to be favoritism based on something like association, rather than actual genetic kinship, sounds like people tying themselves in intellectual knots in order to avoid multi-level selection.

Finally, the EO Wilson quote about “In a group, selfish individuals beat altruistic individuals” Is not a necessary part of multilevel selection theory as I understand it, and indeed may not even be a mainstream position in multi-level selection theory. Certainly, it is easy to argue that in environments where individuals punish bad behavior against them – as all social mammals do – evolutionary game theory shows that altruism readily can spontaneously appear and sustain itself as in tit for tat and more sophisticated direct reciprocity strategies. I expect the differences of opinion on the EO Wilson quote are due to different understandings of punishment of bad behavior and how ubiquitous a part of the environment for group living mammals it is.

While I expect multilevel selection can be robustly defended even with Pinker’s ground-rule of only “natural selection” examples, an alternative response would be to point out the crippled straw horse version of multi-level selection he is arguing against.

Peter Turchin

Mark, I entirely agree. Rather than fight the fight according to the ground rules set by Pinker, we should change the rules. In fact, that is what I do in my commentary. As I just wrote in an e-mail to David Wilson, we can point out that Pinker is not a specialist in this area, and we can write detailed critiques of his reasoning (there is a lot to criticize), but ultimately cultural MLS is going to win not on the strength of such critiques, but by showing that it is a productive research program.


Kin selection acting on memes is simply an obvious application of the theory. In the development of evolutionary theory, group selection explanations come first, and are then obliterated by the far more practical and useful kin selection methodologies. Group selection is OK in theory, but in practice it seems to lack family values.

Mark Sloan

That is, straw man argument, rather than straw horse!

Peter Turchin

Actually, I kinda like the straw horse metaphor!

  1. Home
  2. /
  3. Cliodynamica
  4. /
  5. Regular Posts
  6. /
  7. Steven Pinker on “The...

© Peter Turchin 2023 All rights reserved

Privacy Policy