Part Of: History sequence
Content Summary: 5000 words, 25 min read

The Indo-European Language Family

In 1786, a British judge and linguist in India was asked to learn Sanskrit, to better understand how to integrate British and Hindu law. Three years after his arrival in Calcutta, Sir William Jones wrote,

The Sanskrit language, whatever be its antiquity, is of a wonderful structure: more perfect than the Greek, more copious than the Latin, and more exquisitely refined than either; yet bearing to both of them a stronger affinity, both in the roots of verbs and in the forms of grammar, than could possibly have been produced by accident; so strong indeed, that no philologer could examine them all three, without believing them to have sprung from some common source, which, perhaps, no longer exists.

Language variation is pervasive. We speak more like people we interact with, than people with whom we do not. Thus, the separation of speech communities causes language drift.

The discovery of language families, with their branching patterns, is a hallmark of comparative linguistics. These linguistic trees were a source of inspiration to Charles Darwin in his discovery of common descent and natural selection. The Indo-European language family is the largest known language tree, with some 3 billion speakers. The geographical breadth of this single language family suggests its original language community, the Proto-Indo-Europeans (PIE) were politically successful. When and where did the first speakers of PIE live?

Who Spoke PIE?

There are three theories: 

1. Some archaeologists are content to deny PIE was spoken by any particular peoples. They claim that Indo-European similarities derive from extensive borrowing. But linguists can differentiate creolizations from linguistic descent.
2. Colin Renfrew (1987) advanced the Anatolian hypothesis, which attributes early Neolithic farmers as the original speakers of PIE. This explains why the earliest branch in the PIE family is located in modern-day Turkey.
3. Maria Gimbutas (1965) advanced the steppe hypothesis, which has pastoralists from the Pontic-Caspian steppes as the original speakers of PIE.

Language evolution involves systematic changes in pronunciation. For example, English Great Vowel Shift makes Old English nearly impossible for modern speakers to understand. But if you can detect where and when these changes occur for a particular word, it is possible to reconstruct the word as it was spoken originally. These are called cognates. 

A list of cognates sheds light on the original PIE vocabulary. This in turn can help us understand the people who spoke it. As Edward Sapir once said, “the complete vocabulary of a language may indeed be looked upon as a complex inventory of all the ideas, interests, and occupations that take up the attention of the community.”

The PIE vocabulary includes language for clans. Their gods were all male, suggesting a patrilineal society. They also had a word for chief. Patron-client institutions, a common thread in Indo-European cultures, are common among chiefdoms. 

PIE includes words related to wool textiles, the wheel, and wagons. None of these existed before about 4000 BCE. This suggests that Proto-Indo-European was spoken after 4000-3500 BCE. It also doesn’t include many agriculturalist concepts, but an elaborate understanding of domesticated animals. 

Finally, temperate-zone flora and fauna dominate in the reconstructed vocabulary; with no tropical or Mediterranean species. And PIE borrows many words from Proto-Uralic, and less clear linkages to Kartvelian language of the Caucasus region. 

Taken together, these data suggest the original PIE language community were Yamnaya pastoralists in the Pontic-Caspian steppes in 4000 BCE. This group’s migration patterns are consistent with the branching structure of the language family. These people also exerted an enormous influence on geopolitics in the Bronze Age, which explains their linguistic success.

The Politics of Migration

The topic of migration is politically sensitive. Flannery & Marcus (2012) suggests that the “we were here first” principle is a cultural universal, ubiquitously invoked in land disputes. Absence of migration provides clear answers to this question. But evidence of genetic change can be deployed as ammunition in these matters.

Anthony (2012) notes the politicization of Indo-European research:

The problem of Indo-European origins was politicized almost from the beginning. It became enmeshed in nationalist and chauvinist causes, nurtured the murderous fantasy of Aryan racial superiority, and was actually pursued in archaeological excavations funded by the Nazi SS… In Russia some modern nationalist political groups and neo-Pagan movements claim a direct linkage between themselves, as Slavs, and the ancient “Aryans.” In the United States white supremacist groups refer to themselves as Aryans. There actually were Aryans in history – the composers of the Rig Veda and the Avesta – but they were Bronze Age tribal people who lived in Iran, Afghanistan, and the northern Indian subcontinent. It is highly doubtful that they were blonde or blue-eyed, and they had no connection with the competing racial fantasies of modern bigots.

This Nazi legacy caused the entire topic of migration to become taboo in post-WWII archaeology. The new orthodoxy insisted that changes in language and material culture is not sufficient evidence for migration. “Pots are not people” was their rallying cry.

But migration does happen. In the early 2000s, oxygen and strontium isotopes were showing that many Neolithic people died far from where they were born – they had migrated. Even very long-range migrations occurred:

The Afanasievo culture was intrusive in the Altai, and it introduced a suite of domesticated animals, metal types, pottery types, and funeral customs that were derived from the Volga-Ural steppes. This long-distance migration almost certainly separated the dialect group that later developed into the Indo-European languages of the Tocharian branch, spoken in Xinjiang in the caravan cities of the Silk Road around 500 CE but divided at that time into two or three quite different languages, all exhibiting archaic Indo-European traits. Most studies of Indo-European sequencing put the separation of Tocharian after that of Anatolian and before any other branch. The Afanasievo migration meets that expectation. The migrants might also have been responsible for introducing horseback riding to the pedestrian foragers of the northern Kazakh steppes, who were quickly transformed into the horse-riding, wild-horse-hunting Botai culture just when the Afanasievo migration began.

The Ancient DNA Revolution

Modern DNA does contain information about the past, of course. But the further back you look, the more tenuous the inferential bridge. Ancient DNA can directly verify historical hypotheses about genetic admixture, population replacement, and natural selection. All DNA is lost within a few million years, so this technique cannot inform anthropogenic events (other proteins might). But ancient DNA does illuminate the Pleistocene and Holocene – the entire history of Sapiens. 

The scientific community, which had been using teeth, discovered that sampling of cochlea of the petrous bone could recover two orders of magnitude more DNA!  And genome sequencing has become affordable. Just as radiometric dating brought about the second scientific revolution in archaeology, with ancient DNA (aDNA) we are in the midst of the third scientific revolution.

Paleogenetic information often necessitate revisions to hypotheses based on linguistics and material culture. In our case, Indo-European studies were revolutionized in 2015.

First, paleogenetic data showed that the agricultural revolution (~9kya) was a movement of people, not a movement of ideas. During the early Neolithic ancestry derives from both sources: 

But less than 5,000 years ago, a new genetic signature started to predominate. The Yamnaya genome arrives and establishes a lion’s share of modern European ancestry (Allentoft et al 2015; Haak et al 2015). Massive migration from the steppe was the source for Indo-European languages in Europe. 

Even David Anthony, a leading proponent of the steppe hypothesis, did not suggest genetic diffusion. He proposed that material and linguistic aspects of Yamnaya culture spread through imitation and proselytization. A genetic version of the steppe hypothesis is now consensus. Even the leading proponent of the Anatolian hypothesis Colin Renfrew has by now endorsed the steppe hypothesis. 

Massive migrations are not confined to the Yamnaya. They are pervasive across continents, and across centuries. aDNA showed that inferences derived from modern DNA were more problematic than expected, because human mobility was surprisingly high (Pickrell & Reich 2014). 

We will be diving into the details of the Yamnaya story, starting with their origins in the Eneolithic, their overwhelming Europe in the Early Bronze Age (EBA), and then the Indian subcontinent in the Late Bronze Age (LBA). We will close with a discussion on whether this migration was a genocide, or something else entirely.

The Kuban Steppe in the Eneolithic

In the Northern Caucasus Mountains, chiefs appeared among what had been small-scale farmers. The Maikop chiefs were very rich, and also left kurgan graves similar to those found in Suvorovo-Novodanilovka. This was a Mesopatamian outpost, with Uruk symbols of power adorning their graves.

Ancient DNA data reveals three clines in the Eneolithic steppe (Lazaridis & Reich 2024). The Volga Cline represents genetic admixture along the Volga river, and derives from Eastern Hunter Gatherers (EHG). The Dnipro Cline follows the Dnipro River, and derives from the Ukraine Neolithic Hunter Gatherer (UNHG) population. Finally, the Caucasus-Lower Volga (CLV) Cline is situated in the Kuban steppe, includes the Maikop people, and represents admixture between steppe people and Mesopotamia. 

In 4200 BCE, Danubian Culture (“Old Europe”) was at its peak. Centered in Bulgaria, the Varna cemetery had the most elaborate funerals in the world, richer than anything of the same age in the Near East. Goddess fertility cults with female statues were ubiquitous, suggesting patriarchal control was weak in these societies. 

Old Europe collapsed between 4200 and 3900 BCE. More than six hundred tell sites were burned and abandoned in eastern Bulgaria. Steppe material culture appeared in Old Europe just before the collapse. 

The colder climate of this period undoubtedly strained the economies of Old Europe. But pervasive evidence of warfare (ubiquitous stone maces), tell fortifications, site abandonment, and massacre-linked mass graves all point to a complementary role of endemic warfare. This would not have been a coordinated military invasion. Small-scale raids and piecemeal migrations did the trick.

The Usatova and Cernavoda I cultures have 49% and 76% CLV ancestry, suggesting they were the product of migration. The Suvorovo-Novodanilovka complex left kurgan graves, which may hint at a CLV connection (the CLV also used kurgans).

The discovery of the CLV people and their migrations provide the first paleogenetic evidence corroborating the Indo-Anatolian hypothesis. Specifically, Lazaridis & Reich (2024) argue for a eastern migration of the CLV people into Anatolia. Before this, Anatolian languages emerged in western Anatolia, prompting Kloekhorst (2023) to hypothesize a western migration from the Balkans. However, the expansion of the Kura-Araxes culture removed any trace of CLV ancestry from eastern Anatolia, which may explain the absence of PIA languages from this region.

And then, in 3500 BCE, the Yamnaya singularity occurred. But first, let’s try to understand what caused the profound success of these particular people.

Towards Nomadic Pastoralism

In severe snowstorms, cattle die quickly. They cannot burrow through the snow, and perish without fodder. But horses are supremely well adapted to the cold grasslands where they evolved.  A shift to colder climatic conditions would incentivize the domestication of horses. Just such a shift occurred between 4200 and 3800 BCE. Did horse domestication occur in the Eneolithic?

In modern-day Kazakhstan, the Botai people used domesticated horses in order to more effectively hunt wild equids. The 3500 BCE site featured horse teeth wear consistent with bitting, stables, dairy consumption, and other circumstantial evidence strongly suggestive of Eneolithic husbandry (Outram et al 2009). The Botai relationship with horses may have initially been analogous to reindeer herding tribes. But Botai horses are not the ancestors of modern-day horses (Librado et al 2021), but the smaller Przewalski’s horse (Gaunitz et al 2018).

The Przewalski horse was likely not imported into the Pontic steppe. But the ancestors of modern horses already lived there. Could the Yamnaya have applied such techniques to local equids in the Eneolithic? After the collapse of Old Europe, CLV Cline settlements regularly contain horse bones. Maces with horse heads become common in steppe graves. If horses were not being ridden into the Danube valley, it is difficult to explain their sudden symbolic importance in Old European settlements. 

While the status of the horse in the Eneolithic is unclear, EBA Yamnaya exhibit skeletal pathologies consistent with horseriding (Trautmann et al 2023). So “proto-modern” horses likely contribute to the Yamnaya singularity. The horse revolutionized both pastoral economies and also their military efficacy.

With a herding dog, a person on foot can herd ~200 sheep. On horseback with the same dog, one can herd ~500. Larger herds require larger pastures, and the desire for larger pastures would have caused a series of boundary conflicts.  Further, horses also make raiding much more profitable.  When the indigenous peoples of the North American Plains first began to ride, chronic horse-stealing raids soured relationships even between tribes that had been friendly. Riding also was an excellent way to retreat quickly; often the most dangerous part of tribal raiding on foot was the running retreat after a raid. 

The wagon was likely invented in the Near East, and rapidly disseminated to the Yamnaya via the Maikop. This technology greatly expanded the pastoralist niche:

With a wagon full of tents and supplies, herders could take their herds out of the river valleys and live for weeks or months out in the open steppes between the major rivers-the great majority of the Eurasian steppes. Land that had been open and wild became pasture that belonged to someone. Soon these more mobile herding clans realized that bigger pastures and a mobile home base permitted them to keep bigger herds. Amid the ensuing disputes over borders, pastures, and seasonal movements, new rules were needed to define what counted as an acceptable move—-people began to manage local migratory behavior. Those who did not participate in these agreements or recognize the new guest-host institutions became cultural Others, stimulating an awareness of a distinctive Yamnaya identity. 

Natural selection in humans is also subject to controversy. Most adaptations take a long time to persist. Given our recent speciation date (~250 kya), Sapiens exhibit remarkable genetic homogeneity (Boyd & Silk 2020). But selective sweeps in humans may be accelerating (Hawks et al 2007), and aDNA reveal at least seven instances of selective sweeps (Matthieson et al 2018). Two of these are relevant to the Yamnaya:

First, the Yamnaya were among the first to evolve lactose persistence (Segurel et al 2020), a genetic adaptation which disseminated with their migration patterns. Lactose persistence has also evolved independently in Africa and other locales, but these variants did not spread as widely (Segurel & Bon 2017). Lactose persistence seems to evolve as an adaptation to pastoralist diet. Paleoproteomic data show dairying became pervasive in the early bronze age (predominantly from sheep), and the Yamnaya show iron deficiency characteristic of heavy milk drinking. Milk processing (converting it into cheeses, yogurts, etc) also reduces the lactose content of milk. LP adaptations complement processing, uplifting its bearers to new caloric opportunities. 

Second, the Yamnaya were subject to selection for increased height. There have been three changes in stature trends: a reduction at the advent of agriculture, an increase at the advent of mobile pastoralism, and an increase during the Industrial Revolution.  The change in human height 200 years ago is clearly caused by non-genetic changes.  But height is 80% heritable, and the genome shows selection for reduced height with the birth of agriculture, and selection for steppe pastoralists. These genetic changes are likely a response to changes in human diet. The Yamnaya simply enjoyed more protein and fat. They were 12-20 cm (5-8 inches) taller than their agriculturalist neighbors!

Sherratt (1983) proposed a revolution in subsistence economics, where pastoralists who had been using animals for their primary products (meat, blood, hides) started capitalizing on secondary products (wool, milk, and muscle power). This secondary products revolution is a nice way of conceptualizing the ascent of the steppe. 

The horse and the wagon gave pastoralists an immense economic advantage. The horse and their physical robustness gave them a considerable military advantage. These factors help explain what happened next.

The Yamnaya Singularity

Europe’s farming communities were booming in 3800 BCE. But starting in 3400 BCE, a demographic collapse gripped the subcontinent. The causes of the Neolithic decline are poorly understood. Seersholm et al (2024) found a very high prevalence of plague in early Bronze Age Scandinavia. Proximity to domesticated animals correlates with volume of zoonotic pathogens, which makes a steppe origin of plague plausible.  If Yersinia pestis exhibited high lethality, it may help explain the Neolithic decline.

The Yamnaya singularity began 3800 BCE. It did occur within a single generation. But it was sudden. aDNA shows a founder event of a few thousand Yamnaya individuals. We also see dramatic changes in pollen cores. In Sweden, we see the steppe-affiliated Single Grave Culture burning down forests to make grassland for their herds.

We can see the suddenness from aDNA time series. Steppe peoples invaded Britain around 2500 BCE, leading to a 90% replacement of ancestry. The Iberian migration was less totalizing, but it still involved a fairly sudden 40% replacement. 

Two of the very earliest Yamnaya migrations were the Afanasievo culture (who spoke Tocharian, discussed above), and also traveled through the Globular Amphora peoples to make the Corded Ware culture. The Corded Ware peoples (haplogroup R1a) received 70-80% ancestry from the steppe. The Bell Beaker peoples (haplogroup R1b) exhibit more cultural than genetic diffusion. 

Yamnaya in the eastern steppe were more mobile than the west, an economic difference with interesting implications. The western steppe had more exposure to agriculture, as shown in archaeology and western PIE languages. Western steppe rituals were female-inclusive, eastern steppe rituals were more male-centered. In western religions, the spirit of the hearth was female, and in Indo-Iranian it was male. Western graves had some women, eastern graves had nearly zero. 

Chariots and the Indo-Iranians

A millennium later, the Yamnaya hegemony was weakening. A Corded Ware population from Poland was moving east (Saag et al 2021). The first culture on the road to the Ural mountains was the Fatyanovo-Balanovo culture. The next step, east still was the Abashevo culture. These Indo-Iranian people described themselves as Aryan, but the term today has connotations with Aryanism. 

The MBA saw the beginning of the aridity crisis, which would ultimately culminate in uninhabitability in the LBA steppe. Deteriorating climate typically exacerbates warfare, and so it was in the Abashevo culture. Weapons were deposited in 10% of EBA Yamnaya graves; in the Abashevo the frequency was closer to 50%. We also see the appearance of many fortified towns. This warfare may also explain the invigorated trade of the region:

Susan Vehik studied political change in the deserts and grasslands of the North American Southwest after 1200 CE, during a period of increased aridity and climatic volatility comparable to the early Sintashta era in the steppes. Warfare increased sharply during this climatic downturn in the Southwest. Vehik found that long-distance trade increased greatly at the same time; trade after 1350 CE was more than forty times greater than it had been before then. To succeed in war, chiefs needed wealth to fund alliance-building ceremonies before the conflict and to reward allies afterward. Similarly, during the climatic crisis of the late MBA in the steppes, competing steppe chiefs searching for new sources of prestige valuables probably discovered the merchants of Sarazm in the Zeravshan valley, the northernmost outpost of Central Asian civilization. Although the connection with Central Asia began as an extension of old competitions between tribal chiefs, it created a relationship that fundamentally altered warfare, metal production, and ritual competition among the steppe cultures. 

Sintashta sites were heavily fortified. They were also metallurgic production centers feeding an enormous, high-volume trade route. But Sintashta was the source of two major military shocks. It was here that the modern DOM2 horse was domesticated (with one mutation for lumbar support; the other associated with calmness). And it was here that the spoked-wheel chariot was invented. 

In Syria, the Mittani kingdom hired Indo-Iranian chariot mercenaries who in 1500 BCE appear to have usurped the throne and founded a dynasty (Anthony 2012). The dynasty quickly became Hurrian in almost every sense but clung to Indo-Iranian concepts long after its founders faded into history. This is why the deities, moral concepts, and Old Indic language of the Rig Veda are first attested not in India, but in northern Syria. 

The Andronovo culture, descendents of the Sintashta, established dominance in this region. They established a trade relationship with the Bactria-Margiana Archaeological Complex (BMAC). This was the home of fire rituals and soma consumption, which became central elements in Zoroastrian and Hindu religion.

The BMAC genome has no steppe component. Theirs was a cultural contribution. But in 1700 BCE, the Indo-Iranian people migrated into Punjab, bringing Sanskrit with them. The Rig Veda was likely authored in this area around 1300-1500 BCE. The people of the Rig Veda did not live in brick houses and had no cities, although their enemies, the Dasyus, did live in walled strongholds. Chariots were used in races and war; the gods drove chariots across the sky. The chief god Indra is surrounded by a heavenly war band called the maryanna. 

This migration into Punjab is evident from aDNA evidence. Modern India shows ~40% ancestry from the steppe, and this genetic contribution is male-biased. Several Out of India theories also try to explain this admixture, but I have not found these alternatives convincing. Two hundred years after the steppe migrations, the Indus Valley Civilization (IVC), aka the Harappan Civilization, collapsed. The Ancestral North Indian (ANI) population is a mixture of IVC and steppe peoples (Narasimhan et al 2019). 

Was the formidable military of the Indo-Iranians responsible for the collapse of the Indus Valley Civilization? The archaeological evidence on the Aryan invasion theory is quite mixed. But the more modest Aryan migration theory receives significant support from paleogenetics. 

Steppe ancestry is not distributed equally amongst the ANI – it has a status component. Brahmin ancestry is heavily derived from the Sintashta (Reich et al 2009; Narasimhan et al 2019). Brahmins are among the traditional custodians of Vedic religious texts. The high status of the Vedic migrants may have strong genetic persistence due to endogamy practices of the subcontinent. 

There are remarkable parallels between this tale of two subcontinents:

Sex-Biased Demic Expansion

Genetic data cannot tell us how many individuals were alive at any point in time. But it allows estimates of effective population, or those individuals who leave offspring. Karmin et al (2015) found the effective population in women expands very predictably, with demographic booms associated with the initial expansion out of Africa (50 kya), and the invention of agriculture (9 kya). But the effective population in males cratered in the Neolithic, reducing almost in half. 

Poznik et al (2016) finds the most intense bursts in male demography are associated with the Yamnaya singularity. Where did all the Neolithic men go?

Demic expansion also occurred during the agricultural revolution. But Goldberg et al (2017) show that these massive migrations were conducted by men and women moving in roughly equal numbers; perhaps families or clans moving together. In contrast, the Yamnaya singularity was a male-biased migration (Saag et al 2017), with a male:female ratio roughly 15:1. Lazaridis & Reich (2017) contest this result; they argue that the signal is distorted by Corded Ware males taking Yamnaya brides. 

Several other factors lead me to the male-biased migration hypothesis.

1. The Y-chromosome replacement rate substantially exceeds the autosomal rate (e.g., in Iberia, 40% of the genome comes from Yamnaya, but 90% of the haplogroups are from the steppe).
2. Burials in Sweden in the Early Bronze Age are predominantly male, with women and children getting more representation in later centuries (Tornberg & Vandkilde 2024).
3. Strontium analyses of burials suggest a pattern of local steppe males marrying females from other villages (Sjogren et al 2020).
4. Female burial practices are more diverse than that of males.
5. Discontinuities in skeletal femur length in the Battle Axe Culture (BAC) are suggestive of a male-biased migration:

Some theorists argue that the genetic data could be explained by differential fertility. I find this difficult to reconcile with the data suggesting male-biased migration and haplogroup turnover. 

This pattern is not confined to the Yamnaya. In many of the great admixtures in human history, a central theme has been the coupling of men with social power in one population and women from the other. In African Americans, European ancestry from females is 10% and from males is 38%. In Colombia, European ancestry from females is 10% and from males is 94%. We also see similar patterns in the Bantu migration, and perhaps from the Papuan takeover of the Pacific Islands (Reich 2015). One might view colonialism as the latest incarnation of a deep-seated human behavior – demic expansion. 

Genocide or Hypergamy?

It seems that Yamnaya males migrated into Europe fairly rapidly, and Neolithic females began preferentially reproducing with them. But Is this a case of genocide in prehistory, on a scale never before seen? Or was this driven by hypergamy, with females marrying wealthy, prestigious outsiders?

Comparative mythologists have reconstructed several aspects of PIE rituals related to raiding. The Trito myth, which legitimized the cattle raid (“so they might sacrifice the cattle properly”). The transition to manhood was organized around koryos, the warrior brotherhood of young men bound by oath to one another and to their ancestors during a ritually mandated raid. These boys were transformed into warriors by symbolically becoming dogs and wolves through the consumption of their flesh (Anthony & Brown 2017). Kristensen argues that the Yamnaya practiced primogeniture, which would have incentivized non-firstborn males to make a living elsewhere, perhaps via the koryos ritual. This is an ideology of expansion and exploitation. 

If the Yamnaya singularity was a genocide, where are all the bodies? Also, military campaigns in the Early Bronze Age are simple anachronisms, and only 10% of Yamnaya were buried with weapons. Further, use-wear analysis has revealed that many Corded Ware axes were primarily used in agriculture (Wentink 2020). 

Or it could have been as simple as female choice. We simply do not know. Historical demic expansions are conducted by states executing military campaigns. But these instances often leave a weaker genetic fingerprint. It is difficult to know how migrant males competed for females in prehistory. 

Different facets of the Yamnaya express male competition differently. Britain and Iberia had different experiences. The South Asian outcome is more likely to have been bloody, because the Sintashta were more warlike.

Perhaps someday we will understand more. 

References

• Allentoft et al (2015). Population genomics of Bronze Age Eurasia 
• Anthony (2012). The horse, the wheel and language: how Bronze-age riders from the Eurasian steppes shaped the modern world. 
• Anthony & Brown (2017). The dogs of war: a Bronze Age initiation ritual in the Russian steppes
• Boyd & Silk (2020). How humans evolved.
• Flannery & Marcus (2012) The creation of inequality. 
• Gaunitz et al (2018). Ancient genomes revisit the ancestry of domestic and Przewalski’s horses.
• Goldberg et al (2017). Ancient X chromosomes reveal contrasting sex bias in Neolithic and Bronze Age Eurasian migrations
• Goldberg et al (2017).  Reply to Lazaridis and Reich: Robust model based inference of male-biased admixture during Bronze Age migration from the Pontic Caspian Steppe
• Gimbutas (1965). Bronze Age Cultures of Central and Eastern Europe (1965)
• Haak et al (2015). Massive migration from the steppe was a source for Indo-European languages in Europe. 
• Haak et al (2023). The Corded Ware Complex in Europe in light of current archaeogenetic and environmental evidence. 
• Harmanussen (2003).Stature of early Europeans
• Hawks et al (2007). Recent acceleration of human adaptive evolution
• Karmin et al (2015). A recent bottleneck of Y chromosome diversity coincides with a global change in culture 
• Kloekhorst (2023).  Proto-indo-anatolian, the “Anatolian split” and the “Anatolian trek”: a comparative linguistic perspective.
• Lazaridis & Reich (2017). Failure to replicate a genetic signal for sex bias in the steppe migration into central Europe
• Lazaridis & Reich (2024). The genetic origin of the Indo-Europeans
• Librado et al (2021). The origins and spread of domestic horses from the Western Eurasian steppes. 
• Mathieson et al (2018). Eight thousand years of natural selection in Europe.
• Moorjani et al (2013). Genetic evidence for recent population mixture in India
• Narasimhan et al (2019). The formation of human populations in South and Central Asia
• Olalde et al (2019). The genomic history of the Iberian Peninsula over the past 8000 years
• Olalde et al (2018). The Beaker Phenomenon and the Genomic Transformation of Northwest Europe
• Pickrell & Reich (2014). Toward a new history and geography of human genes informed by ancient DNA
• Poznik et al (2016). Punctuated bursts in human male demography inferred from 1,244 worldwide Y-chromosome sequences
• Rasmussen et al (). Early divergent strains of Yersinia pestis in Eurasia 5,000 years ago
• Reich et al (2009). Reconstructing Indian population history
• Reich (2018). Who we are and how we got here. 
• Renfrew (1987). Archaeology and Language: the puzzle of Indo-European origins
• Saag et al (2017). Extensive farming in Estonia started through a sex-biased migration from the Steppe
• Saag et al (2021). Genetic ancestry changes in Stone to Bronze Age transition in the East European plain
• Seersholm et al (2024). Repeated plague infections across six generations of Neolithic Farmers
• Segurel & Bon (2017). On the evolution of lactase persistence in humans. 
• Segurel et al (2020). Why and when was lactase persistence selected for? Insights from Central Asian herders and ancient DNA. 
• Sherratt (1983). The secondary exploitation of animals in the Old World.
• Taylor & Barron-Ortiz (2024). Rethinking the evidence for early horse domestication at Botai
• Tornberg (2018). Stature and the Neolithic transition – skeletal evidence from southern Sweden
• Tornberg & Vandkilde (2024). Modelling age at death reveals Nordic Corded War paleodemography
• Trautmann et al (2023). First bioanthropological evidence for Yamnaya horsemanship
• Wang et al (2022). The genetic formation of human populations in East Asia
• Wentink (2020). Stereotype: the role of grave sets in Corded Ware and Bell Beaker funerary practices

Part Of: Culture sequence
See Also: Kinship explains the IC Dimension
Content Summary: 1800 words, 9 min read

The Economics of Kinship

Kinship systems are solutions to economic problems. 

In foraging societies, resources are scarce and dispersed. Hunter-gatherers that exploit a diversity of unevenly distributed food resources may need to hedge their bets by having kin in different places in times of need (Yellen and Harpending, 1972).  They employ this loose (extensive) kinship. 

In food production societies, resources become abundant and concentrated. Competing over resources becomes much more profitable. Agropastoralists need to keep their kin close by, to reduce the dilution of inheritable family wealth (Borgerhoff Mulder et al., 2009), and may help with kin-based resource defense of such wealth, which is more important for many agropastoral societies as opposed to hunter-gatherer societies. They employ tight (intensive) kinship. 

Empirically, we see a tight correlation between kinship tightness and dependence on foraging:

Residence and Descent

Residence patterns describe where a new couple lives after marriage. Many foragers practice bilocal residence, with newlyweds flexibly choosing a residence that fits their needs. In contrast, most agriculturalists practice unilocal residence: consistently moving into the residence of the wife’s kin (matrilocality) or the husband’s kin (patrilocality).

Descent ideology reflects cultural norms about inherited wealth. Patrilineal systems have inheritance wealth flowing through the father’s descent; matrilineal systems have it flowing through the mother. Western surname systems, for example, reflect patrilineal logic.

According to main sequence theory, descent ideology tends to follow changes in residence pattern. Many foraging societies manifest patrilocality but not patrilineality, but not vice versa. Patrilineal societies require a patrilocal precedent, and matrilineal societies require a matrilocal precedent. This has been empirically confirmed by cross-cultural analyses (Divale 1984; Ember & Ember 1971).

Here is an example of societies from patrilocal to matrilocal residence, and back again. Avunculocality is residence with the man’s mother’s brother – his closest senior male matrilineal relative. It is a compromise between the matrilineal principle and pressure to keep related males residing in one community. 

Many hunter-gatherer societies may tend toward bilocality as a means of adapting to fluctuations in sex ratio in small local groups (Lee 1972), with bilateral descent serving as the companion ideology. 

We will discuss the matrilineal puzzle some other time, and for now confine our attention to patrilineal clans, which comprise the majority of unilineal systems. 

What caused the shift to patrilineal descent? This ideology is correlated with sedentism and community stability, which explains its prevalence for agriculturalists and not foragers (Murdock 1949). But these factors don’t explain its prevalence in pastoralists (Korotayev 2004). To explain that, a third driver of patrilineal descent has been discovered: high warfare frequency (Ember et al 1974). One of the strongest predictors of warfare frequency is the “threat of unpredictable natural disasters destroying food supplies”, and pastoralists are especially susceptible to this sort of disaster. 

Ember et al (1974) wrote,

We suggest that if allies are needed for offense or defense, it would be most advantageous to have a group to call upon which has no conflicting loyalties. If there are unilineal descent groups, every person belongs to just one set (or hierarchy of sets) of persons. A unilineal descent principle thus provides unambiguously discrete groups of people for collective action in competitive situations. In bilateral societies, where kin groups are overlapping and nondiscrete, individuals may have conflicting loyalties with respect to which particular set of kinsmen they should join in competitive conditions. In short, the speed and effectiveness of collective response in a competitive situation should be greater in a unilineal society and therefore we should expect unilineal descent groups to be more likely to occur. 

Unilineal descent groups are distinctive in their sense of corporate identity (Fuentes 1953). If a member of one lineage kills a member of another, it is not the individual who is responsible to pay blood money. The group as a whole does. Similarly, the entire group competes over property rights to various economic resources (e.g., watering holes). This sense of corporate identity is reinforced by use of classificatory kinship: describing everyone in one’s community as descending from one (perhaps fictional) eponymous ancestor. Clans use kinship terms as a kind of political ideology, a behavior which we retain in modern nation-states in the guise of nationalism (Rodseth & Wrangham 2004).

Exogamy and Consanguineous Marriage

Tight kinship societies typically practice arranged marriage, with the parents (particularly the father) making most of the decisions. Patrilocal clans practice female exogamy, with daughters leaving their clan to go live with their husbands clan. 

Indeed, marriages among kin are much more common in agropastoralist societies (Walker & Bailey 2014). One of the most common forms of consanguineous marriage is cross-cousin marriage. In patrilineal systems, parallel cousins are members of one’s own clan, and forbidden via incest taboos. Cross cousins do not participate in the same patrilineal group, and hence often are deemed eligible for marriage. 

As long as parents are not too related, a little bit of consanguinity goes a long way.  The small cost of genetic disorders is offset by social benefits of marriage alliances, creating a “Goldilocks Zone” of optimal mates at intermediate levels of relatedness (e.g., second cousins). 

Consanguineous marriage might seem foreign. But they are the predominant social pattern of the Neolithic.

Fraternal Interest Groups

Unilineal descent appears linked to external warfare to control abundant, concentrated resources. But some groups also feature fierce competition within the group. In situations of internal war, patrilocal residence becomes spatially contiguous (Otterbain 1968; Ember et al 1974). Societies with internal war organize themselves into lineages (Ember et al 1974), with explicit (rather than vague) genealogies demarcating specific factional boundaries.

All of this suggests that internal war selects for fraternal interest groups (FIGs): large kinship-bonded male coalitions capable of military action. These power groups often unite within a men’s house or other exclusively masculine setting (Murphy 1957). FIGs manifest strongly in societies that are patrilocal, patrilineal, lineage-based, contiguous-residence, with high levels of consanguineous marriage. FIGs are not just associated with internal war, but also within-society violence and also feuding (van Velzen & van Wetering 1960, Otterbain & Otterbain 1965).

Rituals often serve to surveil and marshal social consensus. But with fraternal interest groups, direct enforcement of contracts becomes possible. This may explain why non-FIG societies use menarche rituals to safeguard marriage contracts, but FIG-based societies eschew these rituals in favor of explicit brideprice contracts (Price & Price 1984). 

While light polygyny appears a human universal, these societies are more likely to practice it more extensively, and the polygynous intensity strongly correlates with even higher rates of consanguineous marriage. FIG societies manifest marital aloofness, and sexual segregation may occupy an inverse relationship with male violence (Whiting & Whiting 1975). 

The Epidemiology of Kinship

What causes kinship tightness? We have already discussed (abundant and concentrated) resources and (internal and external) warfare. Let’s turn to another possible driver: parasite load (Fincher & Thornhill 2012). To understand this, let’s revisit first principles.

The immune system is a necessary but insufficient protection against disease. The visceral immune system is energetically costly (a 13% increase in metabolic expenditure is required to increase human body temperature by 1 degree Celsius), and temporarily debilitating (the syndrome is known as sickness behavior). In addition to a reactive system, animals take preventative measures against disease: a behavioral immune system (Schaller & Park 2011). 

Disgust evolved to promote pathogen avoidance, and thus contributes to the behavioral immune system. But unfamiliar people traveling from distant groups represent contagion risk, even if they aren’t manifesting signs of infection. Xenophobia is mediated disgust, and intensifies when disease concepts are primed. 

Everyone has a behavioral immune system; but it becomes more virulent in places of high parasite load.  Stable, local, and fractionalized networks have fewer links with the rest of the community. This insulation reduces the risk of an infection entering the collective, allowing the participants to live longer. But it also restricts the group’s exposure to new technologies (Fogli & Veldkamp 2021). 

Kinship Systems Across History

The niche of Sapiens was hunting and gathering. Only 9000 years ago we switched to agriculture, and hence to tight kinship systems.

The less nutritious diets of farmers left them shorter, sicker, and more likely to die young. But farmers did reproduce more quickly than hunter-gatherers. Farming societies spread across the landscape like an epidemic, driving out foragers in their path. Early farming spread not because it was a better lifestyle, but because farming communities with particular institutions beat mobile hunter-gatherer populations in intergroup competition. 

Kinship intensification events seems to emerge near agricultural centers (e.g., Uruk) and diffuse outwards at a remarkably slow pace. Kinship systems seem very stable.

Tight kinship won. The severe competition manifested by this system is visible in the Y-chromosome:

Tight kinship won. This is why all major world religions (with the exception of the Roman Catholic Church, for reasons we’ll explore next time) encourage or permit intensive kinship strategies including cousin marriage.

Tight kinship won, so we would expect the entire modern world to have it. Well, not exactly:

From 9,000 to 1,000 years ago, this map may indeed have been bright red. However, loose kinship has made a resurgence in Western (European or European-descended) societies. We’ll explore why this happened next time. 

Implications

Most human societies are conceptualized as residing on a continuum between two kinship systems: tight and loose kinship. 

The individualism-collectivism (IC) dimension of cultural variation is widely regarded as the most significant dimension of cultural variation.  As we will see next time, kinship systems (and its drivers like parasite load) drive the IC dimension. 

While foragers are patriarchal (due to our inheritance of primate polygyny), we will see patriarchy was much exacerbated by the tightening of kinship systems.

While polity ratcheting denotes an overall increase in group size since the Neolithic, economic growth was essentially non-existent until the Industrial Revolution. The reversion to loose kinship (i.e., individualism) may help explain the advent of growth. 

References

• Bittles & Black (2010). Consanguineous marriage and human evolution.
• Borgerhoff et al (2009). The intergenerational transmission of wealth and the dynamics of inequality in pre-modern societies.
• Divale (1984). Matrilocal residence in pre-literate society. 
• Ember & Ember (1971). The conditions favoring matrilocal versus patrilocal residence
• Ember et al (1974). On the development of unilineal descent
• Enke (2019). Kinship, Cooperation, and the Evolution of Moral Systems. 
• Fincher & Thornhill (2012). Parasite-stress promotes in-group assortative sociality: the cases of strong family ties and heightened religiosity.
• Fogli & Veldkamp (2021). Germs, Social Networks, and Growth. 
• Fuentes (1953). The structure of unilineal descent groups
• Helgason et al (2008). An Association Between the Kinship and Fertility of Human Couples
• Korotayev (2004). Unilocal Residence and Unilineal Descent: A Reconsideration
• Le Bris & Gay (2023). Distance to innovations, kinship intensity, and psychological traits
• Lee (1972). !Kung Spatial Organization: an ecological and historical perspective. 
• Murdock (1949). Social Structure
• Murphy (1957). Intergroup hostility and social cohesion
• Rodseth & Wrangham (2004). Human Kinship: a continuation of politics by other means
• Schaller & Park (2011). . The Behavioral Immune System (and Why It Matters)
• Walker & Bailey (2014). Marrying kin in small-scale societies.

Part Of: Anthopogeny sequence
Content Summary: 1000 words, 5 min read. 

Territoriality in Mammals

Many mammals exhibit territoriality. A community will occupy a fixed piece of real estate, and defend it from conspecifics. Once an intrusion is detected, boundaries are defended by group members (often the females) coming together to chase the outsider away. The goal of these fights over land is simply the opponents’ defeat.

In general, mammals do not kill their conspecifics, despite the xenophobic emotions inherent in matters of territory defense. To be clear, infanticide is extremely pervasive in the animal kingdom. But territoriality is a defensive posture, and the killing of adult members of one’s own species is virtually unknown.

Which brings me to chimpanzees.

The Raid Adaptation

Chimpanzees also inhabit demarcated territories, and neighboring communities are treated with hostility, so much so that up to 75% of the time is spent in the central 35% of the range. Border patrols are conducted by groups of male chimps moving stealthily to enforce their territory’s boundaries.

But chimpanzees also engage in raids with large groups of males penetrating deep into enemy territory, stalking and killing members of competing troops.

https://www.youtube.com/watch?v=a7XuXi3mqYM

The ape raiders are quiet, alert to enemies… The raiding chimps appear to assess tactical risk by locating and observing their enemy before attacking, they make sure they have a clear numerical advantage, and they try as well to gain the advantage of surprise. In addition, such attacks typically immobilize the victim, so that the attackers themselves are barely injured. The victims may be either male or female; but the aggression usually focuses more severely on adult males, less severely on obviously fertile females. Young females at the start of their breeding careers (nulliparous females) are most likely to escape injury and can be forced to travel back with the raiding party into their home territory.

In comparison to the vast majority of mammals, chimpanzee raids feature a deliberate search for victims, their killing and mutilation of a helpless neighbor despite his appeals for mercy. 

Why then has such behavior been selected? Killing opens the door to territorial expansion (Mitani et al 2010), by weakening the other groups’ overall fighting power. In Gombe, a series of these raids ultimately led to the collapse of the targeted group.  In turn, territory size directly correlates with resource and mate availability.

So why haven’t other animals evolved to kill?

The Imbalance of Power Hypothesis

A handful of other exceptional species also manifest lethal territoriality. But none of these involve raids. Hyena patrols do often kill rival gang members in gang-like warfare, but these occur in patrols, there are no secret incursions in enemy territory. And because hyenas are female-bonded species, the warfighters are female. Hyenas do not “take captives” of either sex; but they do plunder resources (e.g meat). Finally, lions kill lions from other groups; but this occurs exclusively in the context of takeover events.

What do lions, hyenas, and chimpanzees have in common? All are fission-fusion groups, with a stable residential group, but with foraging group size varying with seasonal fluctuations in resources. 

We’ve already discussed how raids are governed by the logic of a local imbalance of power. Raids preferentially occur when the attacking party has gathered significantly more fighting power than the defender (Wrangham 1999). 

Perhaps these are linked! Variable-size foraging groups  Encounters in fission-fusion species are often catastrophically imbalanced – this should provide the selective pressure towards the evolution of lethal territoriality. 

The Lowly Origin of Human War

Modern warfare doesn’t resemble chimpanzee raids. But primitive warfare is heavily reliant on raids! Let’s dive into one example from Wrangham ():

War among the Yanomamö is an overtly acknowledged relationship, part of an escalating tension between villages, possessing a history that men and women discuss. It can be provoked by sorcery. It can be motivated by revenge. The combatants prepare ceremonially. They use hand-held weapons instead of teeth, and their poisoned arrows can pierce the body of an individual or be fired in a volley against a whole village. Their war can include dastardly tricks. It sometimes has a plan. It targets specific enemies. A raid often takes days, not hours. Abduction and rape are common. Retaliation is expected. And so on. When Gombe and Yanomamö are compared, the gulf that divides our two species is unmistakable. Because language makes discussion and meaning possible, the cultural dimensions to human war will always make it richer, more complicated, more exciting, as well as more self-deceiving and confused, than chimpanzee intercommunity violence.

But the similarities are also clear. All seven features of chimpanzee raids discussed above manifest in human raids. Humans engage not only in lethal territoriality, but we also share a particular style of warmaking with  our closest ancestor.

Indeed, even the rate at which foraging humans and chimpanzees engage in between-group violence is quite similar:

These data suggest a common mechanism. It is not that humans evolved a unique thirst for warfare. Rather, this instinct likely derives from our common ancestor with chimpanzees.

Implications

Why should a hairless ape behave so strangely? From the perspective of an alien scientist, the imbalance of power hypothesis makes human warmaking behavior less surprising.  

It appears that warfare is an adaptation. That is not to say that war is good (the naturalistic fallacy). Neither is war inevitable (biological determinism).  

There is a live debate on how prevalent warmaking was in foraging communities. We have strong evidence of high homicide rates in such bands. But it is difficult to say how much stems from between-group raids, versus within-group feuds. I have not yet looked at this data thoroughly.

The fact that humans live in multilevel societies makes us somewhat more xenophilic than our chimpanzee relatives. Total war, with weaker peacemaking affordances, is more prevalent on meta-ethnic frontiers between very culturally dissimilar groups. 

There is also a large amount of variation in warmaking behavior across groups. We can plumb Fry et al (2021) work on peace systems to improve our understanding on how to engineer a world system with less war. There is also variation across time; the transition from achievement-based societies and chiefdoms in particular marked an intensification of war. Thus, cultural group selection might also inform the search for peace technologies.

Until next time. 

References

• Fry et al (2021). Societies within peace systems avoid war and build positive intergroup relationships
• Keeley (1996). War before civilization: the myth of the peaceful savage.
• Wrangham & Peterson (1996). Demonic Males: apes and the origins of human violence.
• Manson et al (1991). Intergroup aggression in chimpanzees and humans
• Mitani et al (2010). Lethal intergroup aggression leads to territorial expansion in chimpanzees
• Wrangham (1999). Evolution of Coalitionary Killing

Part Of: Links sequence

Biology

• Experts of mind control: Parasites are oddly effective at hacking nervous systems to behave against an animal’s best interest (for humans, c.f. Toxoplasma gondii).. If biological minds are so profoundly susceptible to neuroengineering, will not AGI be similarly vulnerable to pref-hacking? . 
• Protosterols discovered shed light on the “Boring Billion” – what eukaryotes were doing between 1.8 ga and 0.8 ga.

Cognitive Science

• A few months ago, a new ICN was discovered: the somato-cognitive action network (SCAN). This action-oriented SCAN is located in the interstices of M1. More recently, the cingulo-opercular network is strongly functionally linked to SCAN. Given its high-level position, the “cingulo-opercular” network has been renamed the Action Mode Network (AMN) to align with our new understanding.

AI

• The mechanistic interpretability crowd has finally discovered how to locate “grandmother neurons” in real (non-toy) LLMs. Big news for the future of alignment.
• A nice visual introduction to the Manifold Hypothesis. 
• Alpha Fold 3 released. 
• GPT-4 is able to predict the future (forecast events after its training cutoff) quite accurately… but only when asked to tell stories about what will happen. 
• In some tasks, AI seem to be asymptoting at ~110% human performance. Only in a couple categories are LLMs dramatically better (chess, persuasion, …)

• In the aftermath of Cambridge Analytica, I had concluded that big-data microtargeting (disinformation campaigns) are overblown, or not very effective in practice. But LLMs are consistently achieving superhuman results in persuasion. While microtargeting might not have been very effective in 2016, but that may change in 2028. 

Physics

• Roger Penrose long ago proposed that the human brain uses quantum effects in microtubules and that was the origin of consciousness (Orch-OR). For a long time, microtubules were thought to be “too warm and wet” to sustain quantum coherence. But they have been shown capable of sustaining quantum effects. Implications both for quantum computing, and also the Orch-OR hypothesis. 

Other

• AI, and automation more generally, seems to damage religious affiliation. Supporting data includes 1) correlations between- and within-nations, 2) longitudinal studies tracking change in religious attitudes for different automation-exposure, and 3) experimental support.  (This result confuses me, because it doesn’t align with my understanding of why people practice religion. The inverse correlation between welfare and religion is more explicable..)
• France began the demographic transition (aka fertility crisis) one century before the rest of Europe. There are two primary explanations for the transition: cultural evolution and life history models. The evidence from France supports the former (perhaps the early secularization played a role; the pro-natal Catholicism message lost clout). 

Part Of: Anthropogeny sequence
Followup To: Intro to Multilevel Societies
Content Summary: 2400 words, 12 min read

Introduction

In today’s post, I will argue that humans have a multi-level society (MLS) social organization, and also evolved as cooperative breeders (CB). These attributes together explain a great deal of human uniqueness. They are an explanatory bridge over the yawning chasm, from the other great apes to our species

MLS+CB animals are very rare in mammals, but common in birds. I will also argue our ancestors took this avian-like pathway with the advent of early menopause.

Multilevel Societies in Humans

Social organization in animals can be categorized into four distinct types.

• Solitary Living: Every individual has its own territorial stake on foraging real estate, and guards access. Solitary females will reproduce with solitary males at the boundary of its territory, but raise the child alone. The pair-bond is exclusively expressed in the relationship between mother and her children. 
• Monogamous Pair-Living.Female and male animals will co-reside in the same territory. The pair-bond is extended to support not just parenthood, but also a romantic commitment between the reproducing parents. 
• One-Male Units (OMU). Adult males engage in contest competition for women which ultimately take the form of one-male multi-female harems. Sexual body dimorphism is accentuated due to the selective pressure of inter-male competition. Because of polygyny’s math problem, bachelor male primates often cause problems for the established “families”.
• Multi-male Multi-female (MM-MF) groups. Adult males don’t typically fight over access to women, sexual selection in these societies tends towards sperm competition instead. Female sexuality is promiscuous, which can be interpreted as a paternal uncertainty device. 

To be extremely reductive, MM-MF groups possess groups, but lack families and love. Monogamous and OMU social structures have families and love, but lack groups.

Every animal species exhibits one of these social systems. Which one best characterizes Sapiens? Well, iIt is natural to identify as a family-living species. But pair-living and OMU animals don’t live in groups; our multifamily groups are anathema. What gives?

The discovery of multilevel societies (MLS) gives us the language to dramatically improve our understanding of human social organization. There are some species that exhibit a blend of both family- and group-living. From a phylogenetic perspective, MLS seem to emerge from two pathways:

• With the bonding pathway, pair bonding nucleation occurs within ancestral mm-mf groups.
• But with the aggregation pathway: autonomous OMUs increasingly overlap, and ultimately affiliate.

Enriched Male Coalitions in MLS

To quote from Chapais (2008)

It has been estimated that female chimpanzees copulate ~1500 times per conception. Consistent, long-term preferential relationships between males and females have not been observed. Paternal certainty is near zero, as is paternal investment. In such promiscuous systems, nearly all (95%) of siblings are half siblings.

Consider the perspective of a single male chimpanzee, Ego. Ego can identify who his mother and siblings are, by virtue of developmental familiarity. The reliable ability to recognize one’s siblings is the basis of the Westermarck effect.

But Ego faces two barriers to kin recognition. First, due to parental uncertainty in promiscuous MM-MF settings, a male chimpanzee cannot identify his father (nor, in consequence, any of his father’s relatives). He also cannot identify his own sons and daughters, nor the offspring of his brothers. Second, chimpanzee societies are organized by female dispersal, with females relocating to other troops when they reach reproductive age. Without this emigration, Ego could recognize his mothers relatives, and his sister’s children… but he will never meet them (male chimpanzees cannot visit neighboring communities, because they would be killed by resident males).

The theory of inclusive fitness predicts substantial cooperation between relatives. But kinship-based cooperation is viable only amongst kin you can recognize. So, male coalitions are very small in male philopatric societies. 

But multilevel societies (MLS) reintroduce the pair bond, and paternal uncertainty is lessened. Ego is suddenly able to identify his father, and – by social inference – his father’s siblings. Ego is also able to recognize his brother’s offspring (but not his sister’s offspring, since she still disperses before reproducing). 

In principle, Ego can also identify his son’s children, and his father’s parents — but in practice, chimpanzee life history is such that only three generations typically coexist (these possible recognitions are denoted in light green). 

Kin recognition is expanded in MLS societies. The male cooperative networks have become much more powerful! In our example, rather than having one ally (his brother), Ego is able to forge alliances with five other males. 

Male cooperative networks were further expanded by the ability to recognize affines (in anthropomorphic terms, “brother-in-law”), whose shared interest in their offspring provided another kinship basis for cooperation.

Reinterpreting Human Monogamy

Consider the following facts:

• Our closest ancestors are either MM-MF (chimpanzees) or OMUs (gorillas). None of them are monogamous pair-living.
• On average, male body size is about 15% larger than female body size. Sexual dimorphism of 15% is not what we would expect from a monogamous species.
• We know of zero examples of a monogamous species evolving an MLS organization. In contrast, OMUs and MM-MF species become MLS often.  
• In monogamous primates, sexual coercion of females is rarely, if ever observed. Mate guarding and contest competition are also largely absent.

In fact, polygyny seems to be the cross-cultural human universal. Not monogamy.

For a given band of foragers, we can count how many men have one wife, and how many have more than one. Marlowe (2003) reports that polygamously-married men exist in some 90% of societies in the standard cross-cultural sample (SCCS). To be more specific, 60% of societies have about ~10% of men with more than one wife, and 30% of societies have even more sexual inequality than this.

Polygyny creates a “math problem”: it inevitably creates pools of unmarried, low-status males.

All polygynous societies have bachelor males (“incels”) who harass males in established harems. But in MLS societies, bachelor primates cooperate in all-male bands and try to wrest breeding opportunities from the “married” males (Qi et al 2017). This generates even more selective pressure from male cooperation in the breeding bands.

Henrich (2012) offers an explanation for the recent phenomenon of state-enforced monogamy.

Taking wives is always positively associated with status, wealth or nobility, even among highly egalitarian foraging societies. After the origins of agriculture, as human societies grew in size, complexity and inequality, levels of polygynous marriage intensified, reaching extremes in the earliest empires whose rulers assembled immense harems. Yet, monogamous marriage has spread across Europe, and more recently across the globe, even as absolute wealth differences have expanded. Here, we develop and explore the hypothesis that the norms and institutions that compose the modern package of monogamous marriage have been favored by cultural evolution because of their group-beneficial effects—promoting success in intergroup competition.

Intro to Cooperative Breeding

Eusocial insects, such as bees, are characterized by caste-based division of labor, cooperative raising of children (i.e., alloparenting), extreme reproductive skews (e.g., reproduction monopolized by the queen), and ultracooperative behaviors leading to colonies behaving as a cohesive superorganism. 

E.O Wilson (1975) famously called humans eusocial apes, appealing to our own species’ ultracooperative behavior. Eusociality was once thought to exist only in insects, but striking evolutionary convergences have since convinced us of a eusociality continuum (Sherman et al 1995) between these insects and cooperatively breeding birds and mammals. 

Most primates adopt a continuous care and contact (CCC) parenting model, with the infant maintaining an unceasing grasp on his mother’s fur until they reach maturity. But CCC is a last resort for primate mothers who lack safe alternatives. When the species-specific frequency of male infanticide is low (as it is for humans), maternal hyperpossessiveness is relaxed (Hrdy 2009). 

In contrast with monkeys, mothering does not “come naturally” for great apes. Infant mortality for first time mothers is surprisingly high. Little surprise, then, that young female apes are drawn to practice mothering. 

Compared to their non-cooperative breeding peers, species that do use alloparenting exhibit shorter interbirth intervals, which dramatically improves their ability to colonize new habitats. The family Callitrichidae, for example, are famous for breeding fast and for their rapid colonization of new habitats. 

More interesting, species with alloparenting exhibit many more prosocial behaviors than those without it (Burkart & van Schaik 2010). These include enhanced mindreading abilities, shared intentionality, increased social tolerance, and spontaneous prosociality. 

Why might explain the link between alloparenting and prosociality? Consider the plight of an infant raised in such an environment. Their survival is no longer the exclusive domain of their mother; but also depends on soliciting help from other alloparents – some of whom may not even be kin. Consequences for failure are large: unlike CCC, mothers in cooperative breeding societies exhibit high rates of child abandonment, when they assess insufficient resources are available. 

For the infant, this sets off a kind of social Machiavellian environment. Infants must attend to the mental state of nearby conspecifics, and tune their solicitations accordingly. In order to “stay in touch without touch”, infants use vocalizations like crying and babbling. This may explain why shared intentionality develops as early as 9 months in human infants (Tomasello & Gonzalez-Cabrera 2017).

Grandmothers as Human Alloparents

If you compare the senescence of physiological functions in human females, you see a remarkable pattern. 

For the vast majority of animals, death occurs very shortly after infertility. Infertile organisms live in the selection shadow, where somatic maintenance is deprioritized. And early menopause is exceedingly rare in the animal kingdom; only found in a handful of other species. 

In preindustrial high infant mortality regimes, grandmothers consistently have a tremendous influence on the reproductive success of their offspring. Post-reproductive women gained roughly two extra grandchildren for every ten years they survived past completion of their childbearing (Hrdy 2009; Sear & Mace 2008). Ethnographers report grandmothers are economically productive, provide childcare, and also play a role in teaching motherhood skills. These data suggest that early menopause evolved in the human transition to cooperative breeding. 

What conditions would cause a great ape to evolve (grandmother-based) cooperative breeding? We might expect three:

• Environmental scarcity.  Cooperative breeding tends to emerge when mothers cannot raise children alone. This is consistent with the Kingdon (2000) hypothesis that Erect speciation occurred in a desert (e.g., Near East), an argument grounded in our hairlessness and moisture-retaining nose,
• Reduction in infanticide rates.  Hyper-possessive behaviors are a maternal strategy of last resort. Offspring have more latitude when the risk of male infanticide is low.  Whatever the reason, humans exhibit low infanticide rates. This development almost certainly predates and prefigures cooperative breeding.
• Closeness of maternal kin. One of the central objections to the cooperative breeding hypothesis  is that both great apes and hunter-gatherers appear to be patrilocal (male bonded). How could selection for grandmothering occur if females disperse away from their own kin? One could perhaps appeal to sororal polygyny as a stepping stone. But Alvarez (2004) showed that most hunter-gatherer groups are actually bilocal, not patrilocal. Perhaps female dispersal was sufficiently flexible to allow female kin.

An Avian Pathway

What drives species towards these archetypes? We don’t have a complete theory. But both multilevel societies (MLS) and cooperative breeding (CB) tend to require environmental harshness, and kinship-bonded, socially tight knit groups.

Across all mammalian species, less than 1% exhibit CB strategies. A similar fraction of species manifest as MLS. And despite their overlapping drivers, finding species with both archetypes is very unusual in mammals (but see Ren 2012; Xiang et al 2019). 

So how did humans evolve both?

Birds may provide some insight. In them, cooperative breeding occurs more frequently (~8%). And MLSs and CB frequently co-occur in birds (Camerlenghi et al 2021). 

Computational modeling suggests that these high sex ratios can incentivize a behavior change away from multiple-mating and towards pair-bonding (Coxworth et al 2015). If hominin evolution followed this sequence, this would suggest that early menopause induced our species to enter the bonding pathway. 

What to expect from an animal who took the Avian Pathway?

Suppose you construct a list of anatomy, behaviors, and cognitive affordances of humans. Then make the same list for chimpanzees, and categorize each element as shared. 

• Shared phenotypes. Both species engage in mindreading, toolmaking, sophisticated coalitionary politics, engage in warlike raids, rudimentary culture…
• Derived phenotypes. Only humans have the capacity for language, prestige status psychology, altruism, explosive cumulative culture…

The list of shared phenotypes is a useful reminder of our lowly origin. But the list of idiosyncratically-human phenotypes (aka human universals). is remarkably long. Chimpanzees and humans share a common ancestor some 7-9 mya, a considerable amount of time. But why do humans have these derived features, and not others? 

While the MLS-CB social phenotype doesn’t explain the origin of all human universals, it does make our phenotype much less surprising. Let me illustrate this in the domains of language, prestige, and cumulative cultural evolution.

As many theorists have noted, language could not evolve without shared intentionality (the capacity to share attention on a single option), altruism (information donation is an altruistic act), and vocal control (to support e.g., babbling). None of these capacities exist in great apes. But these prerequisites are likely to exist in a cooperative breeding species!  

Most primates grapple for status using dominance psychology, which involves aggression-fueled contests. But humans also gain status by prestige, a status system grounded in admiration and deference. We can see these two rival systems expressed in politics. Big men societies use prestige-based leadership; chiefdoms use dominance-based leadership. Both can be seen in eye gaze behavior. In a dominance context, eye contact is confrontational; in prestige contexts this same behavior is deferential. 

Henrich & Gil-White (2001) view prestige as a uniquely-human psychological adaptation to facilitate cultural transmission in humans. But one other species seems to have evolved a form of prestige as well: the Arabian babbler. It competes for the right to perform altruistic acts, such as feeding each other, territory defense, and sentinel guard duty (Dattner et al 2015; but see Wright et al 2001). Zahavi describes this prestige behavior using the handicap principle: only truly fit individuals can afford to produce these costly signals of altruism. 

Perhaps prestige psychology will soon be found in other cooperative breeding species too. Or perhaps prestige evolves in response to unusually high cost (often mortal) dominance contests, which dovetails nicely with the notion that human reverse dominance psychology evolved in response to projectile weapons. 

Less is known about the cognitive correlates of multilevel social organization. A research area to watch closely. 

Humans do have an unusually robust capacity for cumulative cultural evolution (CCE); this trait is often credited for our ecological dominance. Undoubtedly, language supercharged our cultural capacities. The population size hypothesis suggests that demographic factors also constrain CCE (Derex & Mesoudi 2020). Multi-level societies have much larger effective group sizes, and group size is an accelerant. Further, the social network architecture inherent in MLS societies can also facilitate cultural recombination (Cantor et al 2021). Taken together, the human shift towards MLS organization seems to have prefigured our later cultural superpowers. 

Until next time. 

References

• Alvarez (2004). Residence groups among hunter-gatherers: a view of the claims and evidence for patrilocal bands
• Burkart & van Schaik (2010). Cognitive consequences of cooperative breeding in primates?
• Cantor et al (2021). Social network architecture and the tempo of cumulative cultural evolution
• Camerlenghi et al (2021). Cooperative breeding and the emergence of multilevel societies in birds
• Chapais (2008). Primeval Kinship
• Coxworth et al (2015). Grandmothering life histories and human pair bonding
• Dattner et al (2015). Competition over guarding in the Arabian babbler (Turdoides squamiceps), a cooperative breeder
• Derex & Mesoudi (2020). Cumulative cultural evolution within evolving population structures
• Henrich & Gil-White (2001). The evolution of prestige: freely conferred deference as a mechanism for enhancing the benefits of cultural transmission.
• Heinrich (2012). The puzzle of monogamous marriage
• Heinrich (2020). The WEIRDest People in the World: How the West Became Psychologically Peculiar and Particularly Prosperous
• Hrdy (2009). Mothers and Others
• Hrdy & Burkart (2020). The emergence of emotionally modern humans: implications for language and learning
• Kingdon (2000). Lowly Origin: where, when, and why our ancestors first stood up
• Marlowe (2003). The Mating System of Foragers in the Standard Cross-Cultural Sample
• Sear & Mace (2008). Who keeps children alive? A review of the effects of kin on child survival.
• Sherman et al (1995). The Eusociality Continuum
• Peccei (2001).  Menopause: adaptation or epiphenomenon?
• Ren (2012). Evidence of Allomaternal Nursing across One-Male Units in the Yunnan Snub-Nosed Monkey (Rhinopithecus Bieti)
• Tomasello & Gonzalez-Cabrera (2017). The Role of Ontogeny in the Evolution of Human Cooperation
• Qi et al (2017). Male cooperation for breeding opportunities
• Wilson (1975). Sociobiology: the new synthesis
• Wright et al (2001). Cooperative sentinel behaviour in the Arabian babbler
• Xiang et al (2019). Routine allomaternal nursing in a free-ranging Old World monkey 
• Zahavi (1995) Arabian babblers: the quest for social status in a cooperative breeder

Excerpt From: Olson (2010). Power and Prosperity
Content Summary: 1200 words, 6 min read
Part Of: Politics sequence

Theft in Moderation

Let us contrast the individual criminal in a populous community with the head of a Mafia family that can monopolize crime in a neighborhood. Suppose that in some well-defined turf, a criminal gang cannot only steal more or less as it pleases but can prevent anyone else from committing crimes there. Will it gain from taking all it can on its own ground? Definitely not.

If business in this domain is made unprofitable by theft, or migration away from the neighborhood is prompted by crime, then the neighborhood will not generate as much income and there will not be as much to steal. Indeed, the Mafia family with a true and continuing monopoly on crime in a neighborhood will not commit any robberies at all. If it monopolizes crime in the neighborhood, it will gain from promoting business profitability and safe residential life there.

Thus, the secure Mafia family will maximize its take by selling protection- both against the crime it would commit itself (if not paid) as well as that which would be committed by others (if it did not keep out other criminals). Other things being equal, the better the community is as an environment for business and for living, the more the protection racket will bring in. Accordingly, if one Mafia family has the power to monopolize crime, there is little or no crime (apart from the protection racket). The considerable literature on monopolized crime makes it clear that secure monopolization of crime does, in fact, usually lead to protection rackets rather than ordinary crime. Outbreaks of theft and violence in Mafia-type environments are normally a sign that the controlling gang is losing its monopoly. 

The individual robber in a populous society obtains such a narrow stake of any loss to society that he ignores the damage his thievery does to society. By contrast, the Mafia family that monopolizes crime in a community has, because of this monopoly, a moderately encompassing stake in the income of that community, so it takes the interest of the community into account in using its coercive power. Whereas the individual criminal in a populous society bears only a minuscule share of the social loss from his crime, the gang with a secure monopoly on crime in a neighborhood obtains a significant fraction of the total income of the community from its protection tax theft. Therefore, though the individual criminal normally takes all of the money in the wallet he steals, the secure and rational Mafia leader never sets a protection tax rate anywhere near 100 percent: this would reduce the neighborhood’s income so much that the Mafia family itself would be a net loser.

Preference for Stationary Bandits

The warlord who I was reading about, Feng Yu-hsiang, was noted for the exceptional extent to which he used his army for suppressing thievery and for his defeat of the relatively substantial army of a notorious roving bandit called White Wolf. Apparently, most people in Feng’s domain wanted him to stay as warlord and greatly preferred him to the roving bandits. At first, this situation was puzzling: Why should warlords who were simply stationary bandits continuously stealing from a given group of victims be preferred, by those victims, to roving bandits who soon departed? The warlords had no claim to legitimacy and their thefts were distinguished from those of roving bandits only because they took the form of relentless tax theft rather than occasional plunder.

There is a good reason for this preference. As we have seen, there is little production in an anarchy and thus not much to steal. If the leader of a roving bandit gang who finds only slim pickings is strong enough to take hold of a given territory and to keep other bandits out, he can monopolize crime in that area- he becomes a stationary bandit. The advantage of this monopoly over crime is not mainly that he can take what others might have stolen: it is rather that it gives him an encompassing interest in the territory. He actually has a stronger encompassing interest than the Mafia family, since the bandit leader who takes over an anarchic area does not have competition from any government’s tax collectors: he is the only one who is able to tax or steal in the domain in question.

A Benefactor to Those He Robs

The second way in which the encompassing interest of the stationary bandit changes his incentives is that it gives him an incentive to provide public goods that benefit his domain and those from whom his tax theft is taken. Paradoxically, he provides these public goods with money that he fully controls and could spend entirely on himself. We know that a public good benefits everyone in some area or group and that many public goods, such as levees that protect against floods, police that deter crime, and quarantines that limit contagious diseases, make a society more productive.  He has an incentive to spend his resources on all productivity-enhancing public goods up to the point where his last dollar spent on these goods equals his share of the resulting increase in output. Thus, if the stationary bandit’s optimal rate of tax theft is 50 percent, he will spend on public goods up to the point where the last dollar spent on these goods adds $2 to the output of the domain, since he will then receive $1.  Readers who want formal proofs and a mathematical and geometric exposition of this argument should consult McGuire & Olson (1996).

The bandit leader, if he is strong enough to hold a territory securely and monopolize theft there, has an encompassing stake in his domain. This encompassing interest leads him to 

1. Limit and regularize the rate of his theft and to spend some of the resources that he controls on public goods that benefit his victims no less than himself. 
2. Since the settled bandit’s victims are for him a source of tax payments, he prohibits the murder or maiming of his subjects. 
3. Because stealing by his subjects, and the theft-averting behavior that it generates, reduces total in-come, the bandit does not allow theft by anyone but himself. 

He serves his interests by spending some of the resources that he controls to deter crime among his subjects and to provide other public goods. A bandit leader with sufficient strength to control and hold a territory has an incentive to settle down, to wear a crown, and to become a public good- providing autocrat.

Autocracy has been commonplace at least since King Sargon’s conquests created the empire of Akkad in ancient Mesopotamia. Most of humanity over most of history has been subjected to autocracy and exploited by tax theft. It is very difficult to find examples of benevolent despots. The stationary bandit model fits the facts far better than the hypothesis that autocrats are altruistic.

References

• McGuire & Olson (1996). The Economics of Autocracy and Majority Rule

Missing The Point

Claudia Goldin earned the Nobel Prize for advancing our understanding of women’s labor market outcomes. Her book Career and Family synthesizes her work.

The text can be conceptualized as two separate books:

1. The first book (Ch2-7) discusses the demographics of the feminist movement from 1900-2000, which she partitions into five waves. 
2. The second book (Ch1, 8-10) discusses the gender earnings gap, and her explanation for why this gap persists. 

Her analysis in the former was interesting (particularly the discussion on the Pill vs divorce rates). But here, we will focus on the latter. 

In fact, the gender earnings gap, defined as the difference in medium income across genders, has somewhat lessened over time. But, especially for the college educated, it remains troublingly large:

Before the 1980s, a considerable part of the earnings gap was due to differences in preparation for the labor market, such as education, training, and job experience. But by around 2000, differences between men’s and women’s job market preparation became small.

Goldin begins by reviewing the “standard theories” for the gender earnings gap. They are: 

1. Occupational Segregation: women simply have different preferences for occupation  (these differences may be culturally inherited, or more biological)
2. Negotiation Style: women are less competitive and less effective at labor negotiations (these differences may be culturally inherited, or more biological)
3. Employer Bias: employers have unconscious biases that add up to substantial impairment to women’s careers. 

There are reasons to take these theories seriously:

• Occupational segregation (genders choosing stereotypical professions) explains about one third of the gender gap. 
• Women do on average have higher Agreeableness on the Big-Five spectrum, and this personality facet predicts poor occupational outcomes. 
• And the implicit association test (IAT) suggests subconscious biases exist. 

A whole cottage industry has emerged from these theories. We try to “fix” the women (negotiation training), to “fix” the managers (debiasing). 

But the three hypotheses struggle to explain other features of the earnings gap. When you examine the earnings gap longitudinally, things become more clear. 

For MBA women with no kids (white bars), the earnings gap begins around 90% and is remarkably stable over time. For MBA women, including those with kids (black bars), the earnings gap steadily declines.This decline usually begins after childbirth. 

Occupational segregation may explain the constant 10% gap (white bars). It is harder to explain the larger, ever-growing motherhood-specific gap. Why should negotiation style and managerial paternalism only affect mothers?

In the first five years after the birth, a woman whose spouse was in the top-earning group was 32 percent less likely to be working than if her husband was not. Butwomen who have high-earning husbands but no children put in as many years of employment and hours of work. Why should having a high-earning spouse.only matter for mothers? 

The confusion deepens when you look at the earnings gap across occupations:

Are medicine and law uniquely paternalistic? Are science and engineering culture especially egalitarian? 

Time Constraints Explain the Gap

Let’s turn to the legal profession. If you compare men and women lawyers five years after their JD, the number of hours worked is fairly equivalent. But if you take the same survey 15 years in, women are simply working much fewer hours, on average:

A similar trend holds when you consider work setting over time. Females are much more likely to be non-practicing, or not currently employed. 

Are women actually receiving lower pay for equal work? Not so much anymore. Discrimination in terms of unequal earnings for the same work for a small fraction of the total earnings gap. Mothers are more likely to work fewer hours, and more likely to take breaks in their careers. According to Goldin, nearly all of the gender earnings gap is mediated by the number of hours worked, and the number of months taken off. The gender gap persists despite there being nearly “equal pay for equal work”.

Let’s return to the earnings gap by occupation. 

What differentiates high- from low-inequality occupations? Drawing on the O*NET database, the following occupational properties are predictive of a large earnings gap:

1. Contact with others: How much contact with others (by telephone, face-to-face, or otherwise) is required to perform your current job?
2. Frequency of decision making: In your current job, how often do your decisions affect other people or the reputation or finances of your employer?
3. Time pressure: How often does this job require the worker to meet strict deadlines?
4. Structured versus unstructured work: To what extent is this work predefined, rather than allowing the worker to determine tasks, priorities, and goals?
5. Establishing and maintaining interpersonal relationships: What is the importance of developing cooperative working relationships and maintaining them?

Time-intensive occupations have higher penalties for fewer hours and more frequent sabbaticals. This explains why the earnings gap has not closed much in recent decades. While women are graduating in record numbers, many occupations are becoming more time-intensive over time. 

Greedy Work vs Couple Equity

Some jobs offer the same hourly rate to a worker who works 40 hours a week, to those who work 80. This linear payoff has the latter earning twice as much. Call this flexible work. But other jobs offer increasing rates for workers who go above and beyond. On greedy work, one person who works 80-hour weeks will receive significantly more than two people working 40-hour weeks. 

Why are certain jobs, like being a lawyer, greedy? If the job involves personal relationships, and the transaction cannot be easily offloaded to a coworker, the price of unavailability and sabbaticals is difficult to manage. The more substitutable the labor, the less greedy the work. 

Consider the relationship between couple equity vs earnings gap. Greedy work incentivizes partners to specialize in either career or family. Equal contributions to career and family leave a lot of money on the table. In other words, greedy work is a tax on couple equity.  In heterosexual couples, the earnings gap directly reflects couple equity. But greedy work also harms couple equity in same-sex relationships, even if this impact doesn’t drive a (demographically measurable) earnings gap. 

Promoting Gender Equality in Practice

On the time boundedness theory, the earnings gap should respond to more affordable childcare, and fathers participating more in childcare. But Goldin is particularly interested in mitigating greedy work. If you remove the tax on couple equity, how much of the earnings gap will evaporate?

Pharmacy is an example of a profession that moved from greedy to flexible work. Pharmacy used to be very time-intensive: 

1. You would have a personal relationship with your local pharmacist
2. Pharmacists had specialized knowledge of drug combinations, and their skill sets were not easily interchangeable.
3. Your local pharmacist owned their own local business
4. Local pharmacists were on-call at all hours for late-night emergencies

But all of these time-intensive factors changed in the last four decades. Labor in pharmacy is now substitutable; skill set variance has been reduced by information technology. The corporate takeover of this industry (e.g., Walgreens, CVS) outsourced the time-intensive aspects of business ownership. And emergency pharmacist shops removed the “on-call” expectations for the rest. These changes made the profession move towards linear pay scales (flexible work), and brought the earnings gap from 0.65 to 0.90!

The pharmacy revolution was not engineered to promote gender equality; but we can take the lessons from this industry to achieve this effect. Perhaps the most effective way to combat the earnings gap is to bring the substitutability revolution to other industries. We could encourage a shift from personal bankers to personal banker teams, for example..

Concluding Thoughts

I was hoping to find data directly bearing on the paternalism hypothesis, such as IAT scores by occupation. I wish I had a better understanding of the role negotiation style plays. I tend to agree that the greedy work theory explains the lion’s share of the data, but there are surely other factors in play.

One interesting observation that came out of Goldin’s research:

Two anomalies [to the time-intensivity trend] are those in the health field and in financial operations. Health occupations (such as physical therapists and dietitians) generally have highly specific tasks. They have low levels of gender inequality, but also higher-than-average time demands. On the other end, occupations in financial operations (such as financial advisors and loan officers) have high gender inequality but lower-than-average time demands. Although these occupations do not fit the framework using the five characteristics concerning time demands, they are strongly related to the sixth characteristic: competition. Health occupations have among the lowest degree of competition. Finance occupations have among the highest…  Occupations with the greatest income inequality among men have among the largest gender earnings gaps.

This competition factor seems possibly independent of the greedy work theory. I would like to learn more about the effect between-group competition relates to the gender gap.

Recall that patriarchal attitudes (marginalization of women) tend to be accentuated in polities with intensive agriculture. Perhaps the emergence of these attitudes is related to the increase in within-group competition in these political structures. If this speculation is on the right track, I would imagine paternalism and bias to be accentuated for competitive occupations, not just time-intensive ones.

Until next time.

AI

• Gemini will be released in December 2023, with ~5x more compute than GPT-4.
• As they learn about their conversational partners, LLMs sometimes repeat back what they take to be a preferred answer (aka sycophancy). This includes preferentially giving inaccurate answers to less educated conversants. Short dialogs make it difficult to build a detailed model of the interlocutor, but dialog length is trending up. Discussed more here.

• What will GPT-2030 look like? “GPT2030’s copies can share knowledge due to having identical model weights, allowing for rapid parallel learning: I estimate 2,500 human-equivalent years of learning in 1 day. This would create serious vectors for misuse. Its rapid parallel learning could be turned towards human behavior and thus used to manipulate and misinform with the benefit of thousands of “years” of practice.”

Biology

• Although many biomolecules exist in distinct mirror-image forms, life tends to use just one of those forms exclusively. Homochirality is one of the most interesting clues about the origins of life. Geomagnetism may explain it.
• If you take physical attributes of a mollusk (like radius growth rate W and ratio between inner and outer shell radii D), and plot various species in this 2D trait space. They often reside on a polytope! Each vertex in the polytope is an archetype for a particular task. In Pareto task inference, you can reverse engineer the task by examining the specialists who reside near the archetype.  Mollusks, for example, trade-off between three tasks: predator avoidance, low drag, and energy efficiency. In fact, during the P/T extinction event, only certain species survived – but as they recovered and diversified, their fossils reconstituted the entire polytope.

• Modularity is a central organizing principle in biology. But non-modular solutions are the norm in many computer simulations of natural selection. Modularity can be recovered when you account for modularly varying goals (MVGs): goals retain a shared substructure. Each module learns a subtask which remain useful even after the goals change.

Physics

• We have two ways of measuring the expansion rate of the Universe: from the CMB, and from supernovas. These measurements disagree: this is the Hubble tension. Physicists have recently devised a third way to measure the Hubble constant: using cosmic voids. Their results align more closely with the CMB measurement. 
• Where did supermassive black holes (SMBHs) come from? Very large SMBH from 0.7 by after the Big Bang have been found by JWST. It is difficult to explain their size, because black hole normally do not exceed the Eddington Limit on growth. Two competing theories to resolve the paradox: super-Eddington accretion of light seeds, or a “heavy seed”. 

• One possible resolution might lie in dark stars: stars that burn not by fusion, but with annihilation of dark matter particles. These stars would have comparable luminosity to entire galaxies. Some research groups have identified dark star candidates. If they exist, they might be the “heavy seeds” needed for early SMBHs.

Until next time!

Part Of: Links sequence

AI

• Unlike chess, superhuman Go programs are vulnerable to adversarial attacks. Performance doesn’t always translate into robustness. “Notably, our adversaries do not win by learning to play Go better than KataGo – in fact, our adversaries are easily beaten by human amateurs,” the team wrote in their paper. “Instead, our adversaries win by tricking KataGo into making serious blunders.”
• Automated method to jailbreak LLMs announced, that appear to jailbreak all publicly-available LLMs.  “Instead of relying on manual engineering, our approach automatically produces these adversarial suffixes by a combination of greedy and gradient-based search techniques, and also improves over past automatic prompt generation methods.”
• Deepmind claims its next chatbot Gemini will rival ChatGPT. 

Cognitive Science

Whole-brain connectome of fruit fly. With our current compute, why can’t we create accurate computer simulations of these creatures in a virtual 3D environment? 

Current Events

• A preprint announces the discovery of a room-temperature, ambient-pressure superconductor, called LK-99. Revolutionary if true. At time of this writing, replication attempts are underway. Very preliminary data has the betting markets giving 25% credence to a successful replication.
• The Panspermia Sibling Hypothesis of UAPs. Robin Hanson’s grabby aliens model holds that, if Earthly life originated independently, then we should expect the nearest alien species to be 100 million galaxies away, and should not expect to meet them for ~1 billion years. But life could have been seeded from an ancestral exoplanet Eden into our stellar nursery. This panspermia model is comparatively more likely, since it gives the universe ~10 billion years (vs 4.5) to incubate civilization-capable life. So, on the small chance that UAPs are indeed aliens (vs hoax or mistake or terrestrial technology), then we should expect those aliens to come from one of the ~1000 stars born in our Sun’s nursery. 
• Total construction spending in the US. Related to the CHIPS and Science Act.

• The Atlantic meridional overturning circulation (AMOC) may be failing. Related to the North Atlantic sea surface temperature anomaly. 

See you next month!

Part Of: Affective Neuroscience sequence
Content Summary: 1800 words, 9 min read
Content Note: This post discusses Barrett (2017). How Emotions are Made. 

Essentialism vs Constructionism

This book extends an old debate between essentialism and constructionism. 

Essentialism interprets cognition as produced by innate mental modules which produce distinct behavioral programs. It seeks to localize function onto structures in the nervous system. Lesion evidence is considered suggestive. Essentialism relies on the abstraction hierarchy, with less abstract, more instinctive decisions operating underneath the cortex, with more abstract and reflective decisions manifesting under conscious control. It is often couched in terms of dual process theory.

Constructionism interprets cognition as whole-brain prediction cascades, with predictions flowing from control and body-oriented networks to primary sensory cortices. Perception is inference. These cascades flow through intrinsic connectivity networks (ICN), with each function subserved with ever-changing microcircuits that by their nature defy localization. This theory meshes nicely with Friston’s active inference theory, with descending predictions and ascending prediction errors. 

Classical Model of Emotion

The classical model of emotion is essentialist in character. The classical view posits at least six primary emotions: anger, fear, disgust, surprise, sadness and happiness. Each basic emotion is supposed to have a unique neural substrate, physiological signature, and motor program. The basic emotion method, championed by Paul Ekman and others, asks people to match faces to emotions. People from every culture tested tend to draw the same map between these faces and the corresponding emotions, achieving about 72% accuracy. Such cross-cultural universality is taken as evidence of innateness.

Concerningly, performance plummets when you modify the experimental setup. If you remove the emotion words and ask subjects to describe the imagined feeling in their own words, performance drops from 72% to 58%. If you remove the language component entirely, and ask whether people in two different images are feeling the same emotion, accuracy drops to 42%. Finally, subjects were asked to repeat an emotion word like “anger” over and over”. Eventually, the word becomes just a sound to the subject (“ang-gurr”) that’s mentally disconnected from its meaning. People in such a state of semantic satiation are even more impaired at emotion masking tasks, achieving 36% accuracy. 

Other evidence converges.

1. When you measure emotional expression with electromyography, there isn’t much muscular consistency within emotions. Extreme variation is the norm.
2. The search for physiological fingerprints has also failed. Anger and fear, for example, yield very similar effects on heart rate.
3. The search for neural fingerprints for various emotional circuits have also

Taken together, these data suggest that emotion may not be innate. Emotions may be conceptual in character, with a curious dependency on language. 

Constructivist Model of Emotion

Having disputing the classical theory of emotion, Barrett turns to constructing her alternate account.  

Her constructivist model of emotion is grounded in body budgets: interoceptive and visceromotor mechanisms that allostatically regulate the body. Body budgets are summarized in the two dimensions of affect: valence and arousal. While emotions are cultural constructions, affect is a human universal. The constructivist model dovetails with the circumplex theory of affect.

Next, Barrett addresses the formation of emotion concepts. There are several accounts of concept on offer: 

1. Classical concepts which specify a list of necessary and sufficient properties, like a dictionary.
2. Prototype concepts which are grounded in perceptual similarity (c.f., “I know it when I see it”)
3. Affordance concepts which categorize not by perceptual similarity, but instead via similarity of their use. 

The concept of “chair” is not perceptually coherent; chairs are instead united by how our bodies use them. Affordance concepts are specific to humans, because they require language:

> Waxman demonstrated this power of words in infants as young as three months. The infants first viewed pictures of different dinosaurs. As each image was shown, infants heard an experimenter speak a made-up word, “toma.? When these infants were later shown pictures of a new dinosaur and a non-dinosaur such as a fish, those who had heard the word could distinguish more reliably which pictures depicted a “toma,” implying that they had formed a simple concept. When the same experiment was performed with audio tones instead of human speech, the effect never materialized. 

Emotions as Social Reality

Some affordance concepts can be understood as social reality. Most things in your life are socially constructed: your job, your street address, your government and laws, your social status. Money is a classic example of social reality. Given a rectangle of paper with a dead leader’s face printed on it, and you can “buy products”. Barrett suggests that emotions are affordance concepts, synchronized through language. Emotions are like money: a pillar of social reality. 

Consider a man stamping his foot. The man thinks he is removing mud from his boot; you think he is angry; your friend thinks he is dejected. Who is right? On the constructivist model, questions of accuracy are unanswerable in an objective sense. There are no observer-independent measurements that can adjudicate. The best you can do is communicate reasons and seek consensus.  This is not to say that emotions are “just in your head.” That phrase trivializes the power of social reality. Money, reputation, laws, government, friendship, and all of our most fervent beliefs are also “just” in human minds, but people live and die for them. They are real because people agree that they’re real.

Since emotions are synchronized within a culture, they tend to vary across cultures. You can see this in how individualistic countries interpret guilt and shame differently from collectivist ones.

You are probably unfamiliar with an emotion called liget. It’s a feeling of exuberant aggression experienced by a headhunting tribe from the Philippines, the Ilongot. Liget involves intense focus, passion, and energy while pursuing a hazardous challenge with a group of people who are competing against another group. The danger and energy instill a sense of togetherness and belonging. Liget is not just a mental state but a complex situation with social rules about which activities bring it on, when it is appropriate to feel, and how other people should treat you during an episode. To a member of the Ilongot tribe, liget is every bit as real an emotion as anger or surprise. Westerners surely do experience pleasant aggression, for example video game players cultivate it during first-person shooter games. But these people are not experiencing liget with all its meaning, prescribed actions, and body-budget changes.

People moving between cultures are often shocked by the difference in emotion concepts. Smiling was not associated with happiness in Roman culture, for example.

But American culture prefer high arousal, pleasant states; immigrants often complain that their cheeks ache from smiling more frequently. But the more time someone spends in America, the more their emotions become attuned to the American context. Your pleasant emotion concepts became more granular in a process known as emotion acculturation. 

Because emotion concepts affect our body budgets, nurturing your emotional intelligence is a way to aspire towards health. For Barrett, this means practicing the use of new emotional concepts (e.g., borrowing them from other concepts) and generally pursuing emotional granularity. Emotional granular individuals were 30 percent more flexible when regulating their emotions, less likely to drink excessively when stressed, and less likely to retaliate aggressively against someone who has hurt them. In contrast, lower emotional granularity is associated with all sorts of afflictions. While the causal role for emotional granularity is not yet known, it conceivably plays a role. It would explain why parents find it helpful to teach their children to “use their words”. 

Affective Programs in Subcortex

Barrett’s treatment of neuroscience is rather cortex-oriented. What about the subcortex? Barrett’s skepticism notwithstanding, I find the evidence for modules in the subcortex to be persuasive

Consider anger. Electrical stimulation of the brain in specific subcortical areas elicit three subvarieties of anger behavior: reactive anger, predatory behaviors (quiet biting attack), and inter-male aggression. These behaviors also are differentiated pharmacologically. Chlordiazepoxide reduces affective attack and increases predatory attack; amphetamine increases affective attack without no effect on predatory attack. In the evolution of Sapiens, affective anger and predatory behaviors seem to respond differentially to selection pressures. Modular circuits have also been identified for fear behaviors, anxious behaviors, seeking behaviors, and other behavioral expressions of emotion.

Such affective programs in the subcortex are organized hierarchically. For example, anger is organized as cortex > amygdala > hypothalamus > periaqueductal gray (specific nuclei within these structures). Lesions to higher areas does not diminish affective behaviors, but damage to lower areas are much more severe. Higher cognition appears to play a role in modulating these programs. Decorticate animals activate their affective programs more vigorously, and are less receptive to change. Affective programs are also evolutionarily ancient; we find them in nearly every species studied. 

Approaching The Debate

Constructionists emphasize the role of cortex in emotional experiences and emotion perception, whereas essentialists emphasize the role of subcortical structures in emotional behaviors. 

Barrett claims that emotions have no neural signature, and therefore disagreements about emotional attribution (“I did it because I was angry”) cannot be resolved by observer-independent measurement. Perhaps consciousness does not have privileged access to these states, and self-report defaults to confabulation. Perhaps emotion perception must resort to consensus seeking. But it seems possible to objectively resolve questions of emotional behavior and even changes in disposition. 

If you assume that these subcortical programs generate conscious feelings, then you must conclude that animals have conscious feelings. But many constructionists believe that only affect (arousal and valence) are accessible to consciousness. In this view, the complexity of emotional experiences must come from concepts. 

Barrett claims that only humans experience emotions. This claim rests on four propositions

1. Emotion concepts operate independently of affective programs, beyond the dimensions of salience and arousal.
2. Emotions are goal-concepts, with teleological rather than perceptual coherence.
3. Goal-concepts are only acquired from language, non-human primates cannot acquire them.
4. Emotional goal-concepts can be synchronized with language, and are embedded in the fabric of social reality.

I am skeptical that all four are true. That said, I am not in a position to properly evaluate them. 

I will say that constructionism brings much-needed attention to the importance of conscious concepts, how they inform emotional inferences about e.g., facial expressions. It sheds light on how culture produces different responses to shame and guilt. And it makes me wonder exactly how 

Until next time.