EEA baseline
The ancestral social world was structured as a series of hierarchically inclusive, nested relationship layers, each roughly three times the size of the one inside it: an innermost support clique of ~5 close confidants who provided emotional and physical support, a sympathy group of ~15 (people whose loss would be deeply felt, contacted at least monthly), an affinity/band layer of ~50 (the typical overnight foraging camp of casual-but-close associates), and an active personal network of ~150 (the clan/tribal layer, the outer ceiling of stable, individually-known relationships). Each layer carried a qualitatively different grade of intimacy, obligation, and contact frequency, with the heaviest investment concentrated in the inner ~5-15. The ~150 outer figure is an illustrative central tendency, not a constant: traditional tribal and clan groupings cluster in the low hundreds. Its original derivation from the primate neocortex-ratio regression is contested (Lindenfors, Wartel & Lind 2021), so the specific number is treated as approximate; what is load-bearing is the nested layering, not the precise ceiling. This was the full architecture within which status was monitored and alliances tracked: a personally-known reference group bounded at ~150, not an anonymous mass.
EEA value
{
"unit": "individuals",
"layer_sizes": {
"affinity_band": 50,
"active_network": 150,
"support_clique": 5,
"sympathy_group": 15
},
"scaling_factor": 3,
"outer_ceiling_note": "low hundreds; ~150 is an approximate central tendency, not a constant; the neocortex-ratio derivation is contested (Lindenfors 2021)",
"outer_ceiling_approx": 150,
"tribal_clan_grouping": "clusters in the low hundreds (ethnographic)"
}Modern default
The modern environment delivers a radically inverted network shape: nominal friend/follower counts in the hundreds or thousands that vastly exceed the ~150 cognitive ceiling, while the layers that actually carry support and intimacy stay capacity-bounded. Online networks reproduce the same ~5/15 inner structure when genuine ties are measured (support clique ~4, sympathy group ~14, statistically indistinguishable from offline values), and mean platform friend counts hover near ~150-180, but the bulk of nominal connections are inactive labels rather than relationships (only ~27.6% of online 'friends' are reported as genuine). Inner-layer bonds decay back toward mere acquaintance without the face-to-face contact the architecture was built to expect; mediated contact slows but does not substitute for it. The result is a wide, shallow periphery of weak or fictive ties layered on top of a core that the larger network does not, and cannot, enlarge.
The gap
The mismatch is not in the inner-layer capacities themselves (still ~5/15/50/150, not technologically expandable) but between the number of ties the modern environment presents (hundreds to thousands of nominal contacts) and the number the brain can actually monitor and invest in (~150, with real support concentrated in the inner ~5-15). People with unusually large networks do not gain more close friends; they accumulate weakly-defined acquaintances. Status monitoring (M5) and alliance/reputation tracking (M11), evolved for a ~150 personally-known reference group, are now flooded by reference groups orders of magnitude larger and largely composed of strangers, while the trust-bearing inner layers remain capped and are under-fed when face-to-face contact is displaced.
Confidence
This is the one social_structure parameter genuinely full-text-grounded in Cor (Dunbar tradition). The nested ~5/15/50/150 layers, ~3x scaling, and ~150 ceiling are backed by multiple full-text extractions: 404 (Dunbar book, in_physical_collection) and 490 (Social Brain Hypothesis, verbatim Fig.7 quote) for layer structure; 25 (Dunbar book) for the ~150 ceiling and ethnographic mean clan; 483/484/485 (Neocortex 1992 cross-species regression) as the historical cognitive-basis citation; 26 (org-theory 150 threshold). Modern replication of the same architecture and its decay is grounded in genuinely open-access full text (506/507/508/509, Dunbar et al. 2016 R. Soc. Open Sci., DOI 10.1098/rsos.150292, PMC4736918). UPDATE 2026-07-18: the neocortex-ratio derivation of the ~150 figure is now challenged in the corpus (Lindenfors, Wartel & Lind 2021, work 124 / ext 619); contested flipped to true. Grade held replicated for the nested layer STRUCTURE and its comms-data replication, not for the precise number, whose derivation is contested; the precision statistics from the challenged regression (neocortex-predicted ~148, r-squared) have been removed from this parameter's baseline and value. layer_sizes is retained as a structure map (the structure is what survives the challenge). eea_value.outer_ceiling_approx marks ~150 as an approximate central tendency, not a hard interval or constant.
Evidence (10)
- ext#404 · replicated · primary
The innermost group consists of about three to five people.
- ext#490 · replicated · primary
the various human groups that can be identified in any society seem to cluster rather tightly around a series of values (5, 12, 35, 150, 500, and 2,000) with virtually no overlap in the variance around these characteristic values.
- ext#25 · contested · primary
for the twenty-odd tribal societies where census data are available, these clan groups turn out to have a mean size of 153
- ext#483 · cross_species · supporting
Group size is found to be a function of relative neocortical volume, but the ecological variables are not. This is interpreted as evidence in favour of the social intellect theory and against the ecological theories.
- ext#485 · cross_species · supporting
the number of neocortical neurons limits the organism's information-processing capacity and that this then limits the number of relationships that an individual can monitor simultaneously.
- ext#484 · cross_species · supporting
even if it were true that energetic considerations constrain brain size, a proper functional explanation must still be given to explain why brains grow to the size they do.
- ext#26 · theoretical · supporting
organisations of fewer than 150 people work fine on a person-to-person basis, but once they grow larger than this they need a formal hierarchy if they are to work efficiently
- ext#507 · theoretical · supporting
We can only interact coherently with a very small number of other people (about three, in fact) at any one time.
- ext#508 · theoretical · supporting
On average, respondents in Sample 1 considered that only 27.6% of their Facebook friends could be considered 'genuine' (i.e. close) friends, with a strong modal value between 0 and 10%.
- ext#509 · theoretical · supporting
it may be that face-to-face meetings are required from time to time to prevent friendships, in particular, sliding down through the network layers and eventually slipping over the edge of the 150 layer into the category of acquaintances.
EEA baseline
Across a foraging lifetime an individual accumulated a bounded pool of personally known others - people met face-to-face, named, placed in a kin/alliance/reputation map, and remembered. Estimates put this cumulative lifetime roster at roughly 1,000+ distinct individuals (Hill et al. 2014, Hadza/Ache), an order of magnitude above the ~150 actively-maintained network the social brain tracks concurrently. This pool was not a single contemporaneous group: it accreted over decades through camp reshuffling, regional fission-fusion, and durable cross-camp partnerships (e.g. Ju/'hoansi hxaro). Crucially it was finite, slowly-growing, and fully reputation-indexed: nearly everyone one would ever encounter was either already known or one or two reciprocal links away, and almost no one was a true anonymous stranger. The cognitive machinery for naming, tracking reputation, and gauging trust was calibrated to a lifetime ceiling on this order, with the ~150 active-network limit (Dunbar) operating as the concurrent sub-ceiling within it.
EEA value
{
"note": "The ~1000+ lifetime figure is now FULL-TEXT-VERIFIED against its primary, Hill et al. 2014 PLoS ONE (added to Cor 2026-06-09): a forager 'social universe of about a thousand individuals'; lifetime pool of individually-known interactants 'clearly more than 1,000'. It remains a SINGLE-study, model-based estimate on two forager populations (Ache, Hadza) plus a cited San corroboration - illustrative, not a constant. The ~150 concurrent ceiling and nested 5/15/50/150 layers are separately full-text-grounded/replicated (SOC-01).",
"active_network_layers": [
5,
15,
50,
150
],
"concurrent_active_network_ceiling": 150,
"lifetime_distinct_known_individuals": "~1000+"
}Modern default
A networked individual is exposed to vastly more than 1,000 distinct identities - thousands of nominal "friends"/followers plus a near-unbounded daily stream of strangers (crowds, feeds, comment sections, customer interactions). Nominal contact counts routinely run into the thousands while the genuinely-tracked inner network stays pinned near the same ~150, and only ~27.6% of those nominal ties register as genuine close friends. The result is a permanent flood of identities the reputation-tracking and trust-calibration system was never built to index: most are anonymous, non-reciprocal, never re-encountered, and carry no durable reputation, so they neither enter the lifetime roster nor decay out of it normally - they simply accumulate as un-indexable noise. The pool of actually-known, reputation-bearing people is no larger than ancestrally, but it is now embedded in an environment of effectively infinite strangers.
The gap
EEA delivered a finite, slowly-accreting lifetime roster (~1,000+) in which virtually every encountered person was known or one reciprocal link away and carried trackable reputation; modernity delivers an unbounded firehose of identities, the overwhelming majority anonymous, non-reciprocal, and reputation-less. The mismatch is not in the size of the maintainable network (still ~150) but in the ratio of known-to-unknown: a system tuned to "almost everyone you meet is someone you can place" now operates where almost everyone you meet is someone you cannot. This chronically engages strangers as ambient novelty/threat (threat-management framing) and starves the cooperation/reputation system (M11) of the direct, repeated, reciprocal observability it evolved to run on, while the cumulative roster of genuinely-known people fails to grow despite massive nominal connection.
Confidence
GROUNDED 2026-06-09: the defining ~1,000 lifetime-known-individual figure is now full-text-verified against its actual primary, Hill et al. 2014 PLoS ONE (work added this pass; ext H1-H3) - 'social universe of about a thousand individuals' and a lifetime pool 'clearly more than 1,000', which Hill explicitly contrasts with the ~150 'Dunbar number' (attributed to primate brain/group-size regressions). This replaces the prior web/recall, full-text-pending sourcing. Grade set to 'theoretical' (the EEA-mismatch floor): it is a SINGLE study with model-based estimates (yearly interaction rates x life tables, Table S16) on two forager populations, not a multi-study replicated convergence, and the 'lifetime roster conserved while modern stranger-exposure explodes' framing is a mismatch interpretation (capped at theoretical per the hard rule). contested=true retained: EEA-representativeness of Ache/Hadza is genuinely debated and the figure is a single-population model estimate - treat ~1,000 as illustrative. The adjacent ~150 concurrent ceiling / 5-15-50-150 layers remain separately replicated under SOC-01.
Evidence (9)
- ext#502 · replicated · supporting
Each layer increases relationship numbers by an approximate multiple of 3 (5-15-50-150) but decreasing levels of intimacy (strong, medium, and weak ties) and frequency of interaction.
- ext#404 · replicated · supporting
The innermost group consists of about three to five people.
- ext#25 · contested · supporting
for the twenty-odd tribal societies where census data are available, these clan groups turn out to have a mean size of 153
- ext#490 · replicated · supporting
the various human groups that can be identified in any society seem to cluster rather tightly around a series of values (5, 12, 35, 150, 500, and 2,000) with virtually no overlap in the variance around these characteristic values.
- ext#508 · theoretical · challenging
On average, respondents in Sample 1 considered that only 27.6% of their Facebook friends could be considered 'genuine' (i.e. close) friends, with a strong modal value between 0 and 10%.
- ext#509 · theoretical · supporting
it may be that face-to-face meetings are required from time to time to prevent friendships, in particular, sliding down through the network layers and eventually slipping over the edge of the 150 layer into the category of acquaintances.
- ext#606 · thin · primary
Among the Ache and Hadza, frequent visiting and long lifespans mean that adults typically interact with more than three hundred same-sex adults during their lifetimes. This implies a social universe of about a thousand individuals, when opposite-sex adults and children are included.
- ext#607 · thin · supporting
Precontact Northern Ache men, for example, are expected to have hunted with 206 men, conversed with 290 men, and watch tool making by 302 different men in a lifetime (Table S16 in File S1). Eastern Hadza men are expected to converse with 427 other Hadza men, hunt with 303 men, and watch tool making by 395 different men in a lifetime. In contrast, chimpanzees living in typical communities of 11 adult males are expected to interact with a mean of only 21 other adult males in a lifetime
- ext#608 · thin · supporting
Additionally, close companions often interact with a somewhat different set of individuals, so that the total number of indirect interactants that each individual hears about repeatedly in detailed stories, and could expect to possibly meet some time during their lifetime is clearly more than 1,000. This is a much higher number of individually known social interactants than reported for any other primate, and possibly more than any other species on earth. It is also much greater than the predicted 150 significant social interactants (known as “Dunbar's number”) that was extrapolated from primate brain by social group size regressions
EEA baseline
Ancestral humans lived in small, residentially co-present foraging bands (camps): a single overnight camp typically held on the order of a few dozen adults - roughly 25 to 50, with central estimates clustering near ~28-35 - a scale that sits within the Dunbar 'band' layer (~35) and 'overnight camp' layer (~50) of the nested social network. At this scale every co-resident is a personally known, named individual whose behaviour can be directly observed and tracked over time. Band size was thus bounded well below the ~150 active-network ceiling, keeping the day-to-day social field small enough that cooperation, status, and bonding all operated face-to-face among mutually familiar people rather than among anonymous strangers.
EEA value
{
"units": "co-resident adults in a single band / overnight camp",
"source_status": "Full-text-grounded values are the Dunbar layer magnitudes: band ~35 (ext 490, Dunbar 1998) and overnight camp ~50 (ext 404, Dunbar 2010). The 'reported_mean_adults_unverified' 28.2 and the 25-50 spread come from Hill et al. 2011 but are abstract/web-sourced and full-text-PENDING (R2 not satisfied); they carry NO extraction link and must not be treated as established.",
"adult_range_low": 25,
"adult_range_high": 50,
"dunbar_band_layer": 35,
"dunbar_overnight_camp_layer": 50,
"reported_mean_adults_unverified": 28.2
}Modern default
Modern humans do not live in a single bounded, personally-known residential band. Daily life is partitioned across many large, partly anonymous groupings - workplaces, apartment buildings, transit systems, schools, and online networks - most of which exceed the ~25-50 face-to-face band scale and many of which exceed the ~150 ceiling entirely. A person routinely shares physical or digital space with hundreds or thousands of unfamiliar individuals while their actual co-regulating relationships shrink. The small, stable, fully-known co-resident group of a few dozen adults has no direct modern equivalent; its functions are split across institutions where most co-present others are strangers.
The gap
The ancestral social field was a small set of a few dozen (~25-50) personally-known adults sharing a camp, sitting within the band/overnight-camp Dunbar layer. The modern default replaces this with chronic immersion in groupings far larger than the band scale, dominated by unfamiliar people, while genuinely co-regulating relationships number far fewer than the band once held. The delta is a mismatch between the scale at which our cooperation, status-monitoring, and bonding machinery was calibrated (a bounded, fully-known band/camp of a few dozen adults) and the scale we actually inhabit (large, partly anonymous aggregations plus a thin set of disposable ties), so the band-sized layer that once carried daily social life is no longer delivered by any single modern structure.
Confidence
Camp ~30 (range 25-50) now grounded in Smith et al. 2018 full text ("average about 30 individuals"); single Hadza population, cross-society range (Hill 2011 mean 28.2) still contested. contested=true.
Evidence (3)
- ext#490 · replicated · supporting
the various human groups that can be identified in any society seem to cluster rather tightly around a series of values (5, 12, 35, 150, 500, and 2,000) with virtually no overlap in the variance around these characteristic values.
- ext#404 · replicated · supporting
The innermost group consists of about three to five people.
- ext#603 · thin · primary
The Hadza live in temporary camps that average about 30 individuals.
EEA baseline
A person's residential camp/band was NOT a clan of close relatives. Across forager societies the typical residential group (~25-50 people, ~28 adults) was composed mostly of individuals to whom the focal person was not closely related: primary/first-order kin (parents, siblings, offspring) made up only a small minority of co-residents (illustratively under ~10%), and most campmates were unrelated or only distantly related. Because residence was bilocal/sex-egalitarian (couples lived with both the husband's and the wife's kin over time) and camps reshuffled frequently, genetic relatedness within a camp was low. The practical consequence: the durable social glue holding a camp together was NOT shared genes. The bonds that mattered, and that were conserved across the constant churn, were reciprocal, trust-based, reputation-tracked relationships between individuals who chose each other, not kin ties imposed by birth.
EEA value
{
"source_caveat": "figures are Hadza/forager web- and abstract-sourced (Hill et al. 2011 PMID 21393537; Dyble et al. 2015 PMID 25977553); both are real (DOI+PMID) but NOT in Cor and NOT full-text-verified in Cor (retrieval returned is_fulltext=false for each)",
"camp_size_total_range": [
25,
50
],
"camp_size_adults_approx": 28,
"relatedness_qualitative": "low; most campmates unrelated or distantly related",
"primary_kin_fraction_of_camp": "<~0.10"
}Modern default
Modern social environments invert the ancestral pattern in two opposite directions at once. (1) The family one is biologically related to is now small, geographically dispersed, and often the ONLY kin one regularly co-resides with, so the kin one has are concentrated in a tiny stable household rather than diffused through a larger chosen group. (2) The larger groups one spends daily life inside (workplace, neighborhood, online networks, institutions) are composed of effectively unrelated strangers, but unlike the ancestral camp these groupings are NOT self-assembled through mutual choice, reciprocity, and reputation tracking; membership is assigned by employment, address, or platform. The result is that 'who you live and work among' is no longer the output of a relationship-selection process; it is administratively or commercially determined, and the unrelated people around you are frequently strangers you did not choose and do not reciprocally track.
The gap
The ancestral signal was: 'the group around me is mostly unrelated people I (or my band) actively chose and continuously vetted through reciprocity and reputation; kinship is not what holds us together, the relationships are.' The modern signal is degraded on the selection axis: the unrelated people around us are still mostly unrelated (that part matches), but they are assigned rather than chosen, and the reciprocity/reputation machinery that turned unrelated campmates into durable trusted allies is largely absent or shallow. Mechanisms M3 (bonding) and M11 (cooperation/alliance) evolved expecting that low kin-relatedness would be compensated by self-selected, reciprocal, reputation-anchored dyads. Modernity delivers the low relatedness WITHOUT the compensating chosen-and-vetted dyadic structure, so the mechanisms fire in a world of unrelated-but-also-unbonded co-residents, which is a key part of why being surrounded by people can still register as social isolation.
Confidence
Low within-camp relatedness now grounded in two full-text primaries (Apicella 2012: "first-order relatives make up less than 10% of residential camps"; Smith 2018: primary kin 1.43/1.93 per camp). Both single-population / partly restating Hill 2011; kept thin.
Evidence (2)
- ext#601 · thin · supporting
first-order relatives make up less than 10% of residential camps
- ext#605 · thin · primary
Relatedness within camps is low with primary kin comprising, on average, 1.43 and 1.93 of men and women's campmates respectively.
EEA baseline
The residential group (camp/band) was a fission-fusion structure that continuously dissolved and reconstituted rather than a fixed-membership unit. Among the best-studied foragers (Hadza), entire camps relocated roughly every 4-8 weeks and individuals reshuffled across camps such that only a minority of campmates carried over from one year to the next. The one full-text-verified figure is from a multi-year Hadza study (Smith et al. 2018, bioRxiv 313064): a mean proportion of repeated campmates of ~12.1%, implying ~88% annual turnover. Crucially, despite this high residential churn, assortment on cooperation persisted (cooperators clustered together across reshuffles), and what was conserved across the turnover was not the group but the durable, individually-tracked reciprocal relationships between specific people. Group composition was fluid; the dyadic ties were the stable substrate.
EEA value
{
"source": "Smith et al. 2018 (bioRxiv 313064), Hadza, full-text-read in Retrieve phase",
"not_yet_in_cor": true,
"unverified_secondary_note": "A higher figure of ~21.9% repeated / ~78% turnover circulates in secondary references and the task brief, but the Smith 2018 full text does NOT report it (it states 12.1%). That number is mis-attributable to this source and is plausibly from Smith et al. 2016 (R. Soc. Open Sci. 3:160131); it is NOT used as a value here and must be separately verified before any use.",
"annual_turnover_proportion": 0.879,
"camp_relocation_interval_weeks": [
4,
8
],
"repeated_campmates_proportion_per_year": 0.121
}Modern default
Modern social environments invert the structure: the GROUP labels are stable and durable (a workplace, a school cohort, a neighborhood, a social-media platform or feed) while the actual dyadic relationships within them are shallow, interchangeable, and disposable. People remain nominal members of named, persistent groups for years, yet cycle through co-members with little individual reciprocal investment, and there is no institution analogous to a delayed-reciprocity partnership (e.g. hxaro) that conserves a specific bond once physical proximity ends. Relationships routinely lapse the moment shared context (job, class, proximity) is removed.
The gap
The ancestral and modern environments are inverted on which layer is stable. Ancestrally the GROUP churned (~88% of campmates turned over yearly) while the durable reciprocal DYAD was conserved across the churn; the relationship, not the group, was the unit that persisted. Modernity delivers the opposite: stable, durable group labels (workplace, cohort, feed) paired with shallow, disposable dyads and no mechanism (no hxaro-equivalent) for conserving a specific bond once proximity ends. The mismatch is not 'too much turnover' versus 'too little' - it is that turnover now lands on the relationships (which evolved to persist) instead of on group membership (which evolved to be fluid), leaving the co-regulating, reputation-tracking reciprocal dyad chronically unmet.
Resolution unit
The durable reciprocal dyad - a specific, individually-known partner with whom reciprocity, co-regulation, and reputation are tracked over years. What needs conserving across membership turnover is not the group or the amount of contact but the persistence of particular bonded relationships; the bond, not the band, is the conserved unit.
Confidence
Turnover now grounded in Smith et al. 2018 full text: mean repeated campmates 12.1% (=> ~88% annual churn), camps shift every 4-8 weeks. CORRECTED from the brief's 21.9%/78% (that figure traces to Smith et al. 2016, a different paper). Single Hadza population; kept thin.
Evidence (2)
- ext#602 · longitudinal · primary
Across the three intervals, the mean proportion of repeated campmates was 12.1%.
- ext#604 · thin · supporting
Every 4-8 weeks entire camps shift location usually in response to resource availability.
EEA baseline
Who you lived alongside was not assigned by birth or fixed institution but produced by an ongoing, choice-driven assortment process operating across frequent residential reshuffling. In Hadza camps this "algorithm" had recurring structure: (1) homophily - co-residence was more likely between people similar in age, height, weight, body fat, and handgrip strength; (2) strong reciprocity - naming was near-symmetric, with an ego roughly 37-51x (point ~44x) more likely to name an alter who named them; (3) cooperator assortment - cooperators clustered with cooperators, an association extending out to two degrees of separation in gift networks; plus transitivity and geographic decay. Crucially, kinship was NOT the main glue: people actively sorted on disposition and reciprocity, not primarily on relatedness. The output was a camp continually re-composed of preferred, reciprocating, similarly-cooperative partners.
EEA value
{
"homophily_traits": [
"age",
"height",
"weight",
"body_fat",
"handgrip_strength"
],
"kinship_is_primary_glue": false,
"reciprocity_naming_odds_multiplier": {
"ci_95": [
37.6,
51.4
],
"source": "Apicella et al 2012",
"population": "Hadza",
"illustrative": true,
"point_estimate": 44.2
},
"cooperator_assortment_degrees_of_separation": 2
}Modern default
Group membership is overwhelmingly assigned rather than self-assorted: workplaces, school classes, residential buildings, and algorithmically-curated feeds place you among people you did not choose and who did not choose you. Where modern matching exists (dating apps, professional networks, recommender feeds), the optimization target is engagement, novelty, or proximity-of-the-moment - not durable mutual reciprocity, demonstrated cooperativeness, or transitive trust within a personally-known group. The ancestral feedback loop (you preferentially co-reside with people who reciprocate your investment and who others vouch for, observed directly over time) is largely absent; assortment on cooperation is replaced by assortment on platform metrics or institutional convenience.
The gap
The EEA delivered a continuous, agency-rich sorting process that filtered one's immediate social world toward reciprocating, cooperative, vouched-for partners - kinship was secondary. The modern environment inverts this: membership in one's daily groups is mostly imposed (employer, building, algorithm), and the explicit matching systems that do exist optimize for engagement/novelty rather than reciprocity and demonstrated cooperation. The result is reduced control over WHO populates one's social field, weaker assortment on actual cooperativeness, and loss of the transitive-reputation signal that let ancestral humans cluster with trustworthy reciprocators out to two degrees. The mismatch is not "fewer people" but a broken selection rule: the inputs the ancestral algorithm ran on (mutual choice, observed reciprocity, cooperator reputation) are no longer the primary determinants of who one lives and works among.
Confidence
Shuffle algorithm now grounded in Apicella 2012 full text: trait homophily, strong reciprocity (44.2x campmate / 14.3x gift), and social distance as important as genetic relatedness (kinship not the sole glue). Single Hadza population; kept thin.
Evidence (3)
- ext#598 · thin · primary
An "ego" (the naming person) is 44.2 times (95% C.I. 37.6 to 51.4) more likely to name an "alter" (the named person) in the campmate network, and 14.3 times (95% C.I. 12.2 to 16.4) more likely to name an alter in the gift network, if the alter reciprocated the social tie by also naming the ego as a friend
- ext#599 · thin · primary
in both the campmate and gift networks, social ties are significantly more likely when two people are similar in age, height, weight, body fat, and handgrip strength.
- ext#600 · thin · supporting
social distance appears to be as important as genetic relatedness and physical proximity in explaining assortativity in cooperation
EEA baseline
In ancestral foraging life the load-bearing social unit was the durable, reciprocal DYAD, not the group. Because groups continuously reshuffled (camps fissioned/fused on the order of every ~4-8 weeks, with the large majority of campmates turning over year-to-year), what was conserved across the churn was the relationship itself, sustained for decades by both face-to-face co-presence when proximity allowed AND by explicit maintenance institutions when it did not. Among the Ju/'hoansi, hxaro was a formal delayed-reciprocity gift-partnership system reported to begin in childhood, be kept for life, be inherited at death, and span dispersed bands up to ~200 km apart, with only a minority of a person's partners co-resident in the current camp, so a bond could persist even when the two partners were rarely or never in the same camp. Two adults who reached adulthood together could expect on the order of ~27 years of overlapping co-survival, and reciprocal alliances ran over that whole horizon. Trivers's conditions for the evolution of reciprocal altruism describe exactly this regime: a long lifetime, low dispersal, and repeated interaction with the same small set of individuals, so that an investment in a specific partner reliably paid back over time.
EEA value
{
"_note": "All numeric values are ILLUSTRATIVE and full-text-PENDING in Cor; NO extraction id is cited for any of them. The ~27-yr co-survival ceiling and the hxaro figures (childhood-onset, lifelong, inherited, <=200 km, ~18% co-resident) are attributed to Wiessner 2002 EHB 23:407-436, which is NOT in Cor and whose full text could NOT be obtained (retrieval is_fulltext=false, zero verbatim quotes; snippet figures rejected per R2). For turnover: the only retrieval-grounded figure is from Smith et al. bioRxiv 313064 (mean repeated campmates 12.1% => ~88% annual turnover; camps shift every ~4-8 weeks), which is itself NOT yet a Cor extraction; the ~78% lower bound of the range is web/recall-sourced and full-text-UNVERIFIED and must NOT be attributed to bioRxiv 313064. Treat the range edges as soft.",
"camp_reshuffle_interval_weeks": [
4,
8
],
"hxaro_partner_max_distance_km": 200,
"co_survival_ceiling_years_approx": 27,
"annual_campmate_turnover_pct_approx": [
78,
88
],
"hxaro_partners_in_own_camp_pct_approx": 18
}Modern default
The modern environment inverts the ancestral pattern: it supplies stable GROUP labels (a workplace, a school cohort, a social-media feed, a city) while the dyads inside them are shallow and disposable, and there is no institutional equivalent of hxaro that conserves a specific bond once physical proximity ends. When two people stop sharing a context, the relationship is simply allowed to decay; nothing structurally obligates either party to keep servicing it. Empirically, friendships require recurring face-to-face contact to avoid sliding down through the network layers and eventually dropping out of the personal network entirely, and online media at best slows that decay rather than preventing it. Trust and bond depth track time invested in a relationship, and the maintenance cost (mainly time) scales roughly linearly with the number of relationships, so modern people spread thin investment across large, low-trust networks and let most ties lapse rather than maintaining a small set of partners across decades.
The gap
The ancestral environment conserved the relationship across changing groups; the modern environment conserves the group label and lets the relationships churn. The specific missing piece is a maintenance institution: the EEA had hxaro-type structures (formalized, inheritable, distance-spanning, lifelong delayed reciprocity) that actively preserved a named dyad even with little or no co-presence, whereas modernity has no hxaro-equivalent and instead relies on shared proximity/context as the only glue. When that proximity ends (job change, move, leaving school), the bond is not migrated to an institution that keeps it alive; it decays. The result is many stable-looking memberships sitting on top of disposable, shallow ties, with the durable decades-long reciprocal partnership that M3/M11 are calibrated to expect largely absent.
Resolution unit
The durable, spatially-dispersed reciprocal DYAD plus the maintenance institutions that conserve it across distance and time - i.e., a specific named partner whose bond is actively serviced (face-to-face when possible, and via hxaro-type delayed-reciprocity exchange when apart) so the relationship persists for decades and survives the reshuffling of any particular group. Conserved across the churn is the relationship, not the group; the resolution requires migrating a bond to an institution that keeps it alive once proximity ends, not merely accumulating more low-investment ties.
Confidence
Mixed and honestly LOW. The headline EEA numbers for this parameter (~27-yr co-survival ceiling; hxaro childhood-onset, lifelong, inherited, <=200 km, ~18% co-resident) come from Wiessner 2002 (Evolution and Human Behavior 23:407-436, DOI 10.1016/S1090-5138(02)00096-X), which is NOT in Cor and whose full text could not be legitimately obtained (paywalled Elsevier; retrieval returned is_fulltext=false with zero verbatim quotes; snippet figures were rejected per R2). The turnover/reshuffle figures trace to Smith et al. (bioRxiv 313064: 12.1% repeated => ~88% turnover; camps shift every ~4-8 weeks), which is likewise not yet a Cor extraction; the ~78% lower bound is web/recall-sourced and full-text-unverified. So NO extraction id is cited for any specific ancestral number here; those values are flagged ILLUSTRATIVE / full-text-pending. The evidence that IS cited grounds only the general durability-and-maintenance logic and the modern-side decay, from genuine full-text-backed Cor extractions: 509 and 510 (Dunbar 2016, rsos.150292 - face-to-face contact required to prevent friendship decay; media only slows decay), 503 and 505 (Sutcliffe, Dunbar, Binder & Arrow 2012 - maintenance cost is time-linear; trust is a function of time invested), and 479 (Trivers 1971 - reciprocal altruism requires long lifetime, low dispersal, repeated interaction with the same small set; author_quote intentionally absent/None per its remediation, so no verbatim quote is attributed). Grade held at 'thin' because the parameter-defining ancestral figures are unverified and the cited extractions address the why/modern-half rather than the EEA co-survival/hxaro specifics. Not 'replicated' despite some cited extractions being individually replicated, because they do not converge on this parameter's ancestral claims.
Evidence (5)
- ext#509 · theoretical · primary
it may be that face-to-face meetings are required from time to time to prevent friendships, in particular, sliding down through the network layers and eventually slipping over the edge of the 150 layer into the category of acquaintances.
- ext#510 · theoretical · supporting
social media may function mainly to prevent friendships decaying over time in the absence of opportunities for face-to-face contact.
- ext#503 · theoretical · supporting
At each layer, the benefits are asymptotic but the costs of maintaining a relationship at that level (most obviously, the time that has to be invested in servicing it) are roughly linear with the number of relationships.
- ext#505 · theoretical · supporting
psychologically, these trade-offs are related to the level of trust in a relationship, and that this is itself a function of the time invested in the relationship.
- ext#479 · theoretical · supportingTrivers identifies conditions maximizing the chance reciprocal altruism will be selected for: (1) many altruistic situations in the lifetime of the altruists, (2) a given altruist repeatedly interacts with the same small set of individuals, and (3) pairs of altruists are exposed
EEA baseline
Social cognition operated over a bounded pool: a personally-known reference network capped near the ~150 Dunbar ceiling, nested in ~5/15/50/150 layers of decreasing intimacy, the outermost being the people one recognizes and interacts with across a year. Within that architecture almost every face encountered was already known, placed, and reputationally tracked. A genuinely novel stranger - an individual who could not be located anywhere in one's known network or in the known networks of one's allies - was therefore a comparatively rare event. Because the bounded pool had no slot for such a person, novelty itself was diagnostic: a never-before-seen individual could not be vouched for, had no shared reputational history, and could not be assumed to be bound by the local web of reciprocity and norm enforcement. Encounters with unplaceable strangers were thus both infrequent and highly salient, and the threat-management system (M1) treated unfamiliarity as a plausible danger cue to be resolved (placed, vouched-for, or avoided) rather than as a routine background condition. This is an inference from the bounded-pool architecture plus M1, not a directly measured ancestral rate.
Modern default
The contemporary environment delivers a near-continuous stream of novel, unplaceable faces: dense urban crowds, transit, commerce, and especially the feeds and follower graphs of online platforms expose a person to more unfamiliar individuals in a single day than an ancestral forager might meet in a lifetime. Unfamiliarity is the modal condition rather than the rare exception, so the diagnostic value of novelty collapses - a stranger is no longer informative about anything. Yet the same M1 machinery still treats unplaceable others as potential threats, now firing against a constant baseline it was never calibrated for, while the reputational and vouching scaffolding that once resolved a stranger encounter is largely absent.
The gap
The delta is one of base rate and resolvability. Ancestrally, novel strangers were rare, salient, and resolvable (place them in the known web, or treat unfamiliarity as an actionable threat signal); modernity inverts this to constant, non-diagnostic, unresolvable novelty. The threat-management system inherited from the rare-stranger regime cannot down-regulate to the new base rate: it either over-fires (chronic low-grade vigilance, stranger-directed wariness in crowds and online) or, where it habituates, leaves people transacting with a stream of others who carry none of the reputational accountability the system implicitly assumes. The conserved unit being eroded is not the count of strangers but the durable, reputationally-tracked dyad against which any new face used to be evaluated; once almost everyone is a stranger, the bounded-pool premise that made novelty meaningful no longer holds.
Resolution unit
The durable, reputationally-tracked reciprocal dyad: ancestrally every new face was evaluated against a finite web of known, vouched-for relationships, and it was that web of bonds - not the count of strangers per se - that gave novelty its diagnostic and protective meaning. The mismatch resolves only insofar as new encounters can be anchored back into durable, accountable relationships rather than left as anonymous one-off contacts.
Confidence
Pure inference: no Cor extraction and no full-text retrieval quote directly measures or states an ancestral novel-stranger frequency, so eea_value is null and the grade is theoretical. The premise the inference rests on - a bounded, personally-known reference pool capped near ~150 and nested in ~5/15/50/150 layers - IS full-text-grounded via Dunbar's social-brain work: extraction 490 (work 69, DOI 10.1002/(SICI)1520-6505(1998)6:5<178::AID-EVAN5>3.0.CO;2-8; R1 satisfied) on familiarity-based layer clustering, and extraction 485 (work 68, DOI 10.1016/0047-2484(92)90081-J; R1 satisfied) on the neocortical limit to simultaneously-monitorable relationships. Both are cited as SUPPORTING (premise), not primary, because they bound the known pool but do not themselves count strangers. The complementary M1 claim that unfamiliarity functions as an innate danger cue is most directly stated by Bowlby (Attachment and Loss Vol. 2), but that work (id 98) has no DOI/PMID, is not in the physical collection, and its extraction (574) records the book was not directly accessed and carries no verbatim quote (author_quote=None) - it fails R1 and R2 and is deliberately NOT cited; the stranger-as-threat link is full-text-pending. Tomasello extractions 238/250 on cooperation with 'in-group strangers' were considered and rejected as cutting against, not supporting, the rare-stranger-as-threat framing. Not contested as an inference within the framework; it is simply unmeasured.
Evidence (2)
- ext#490 · replicated · supporting
the various human groups that can be identified in any society seem to cluster rather tightly around a series of values (5, 12, 35, 150, 500, and 2,000) with virtually no overlap in the variance around these characteristic values.
- ext#485 · cross_species · supporting
the number of neocortical neurons limits the organism's information-processing capacity and that this then limits the number of relationships that an individual can monitor simultaneously.
EEA baseline
In a band/camp embedded in a personally-known reference group of roughly 150 individuals, virtually all of a person's relevant behavior was directly observable by, or rapidly transmitted through gossip to, people who knew them by name, knew their history, and would interact with them again for decades. Reputation was a shared, continuously-updated, distributed record: as Boehm puts it, "the band keeps a dossier on every individual, noting positive and negative points." Observability had teeth: those who deviated (theft, deception, stinginess, free-riding) were subject to gossip, ridicule, ostracism, and ultimately exclusion, and everyone knew this monitoring was constant, so it functioned as a deterrent and people exercised self-control accordingly. Because foraging was obligately collaborative, individuals had to continuously evaluate partners and track how they themselves were being evaluated. The net result: visibility was near-total, embodied, hard to fake over time, attached to a single durable identity, and backed by real consequences enforced by people one could not escape.
EEA value
{
"identity": "single durable identity, behavior history attached for life",
"observability": "near-total within band/camp; rapid gossip diffusion across the ~150 layer",
"accountability": "high: gossip, ridicule, ostracism, exclusion; non-exitable repeat interaction",
"reference_group_basis": "personally-known, named, repeat-interaction reference group (Dunbar ceiling)",
"reputational_reference_group_persons": 150
}Modern default
Most behavior now occurs among strangers or weak ties who do not know one's history and will never interact with one again, while the visibility that does exist is mediated, curated, and quantitatively inflated. On social platforms only ~27.6% of nominal "friends" are considered genuine close ties (with a strong modal value between 0 and 10%), so networks are padded with weakly-accountable acquaintances rather than people tracking one's actual conduct. Mediated contact cannot substitute for the embodied, face-to-face observation the bonding/reputation system evolved for: without periodic in-person contact, ties decay through the network layers into mere acquaintance regardless of online activity. Visibility is now selectively self-presented (gameable), fragmented across multiple personas and contexts, frequently pseudonymous or low-stakes, and policed (if at all) by distant institutions or algorithms rather than by named peers with whom one shares a future. Reputation is either absent (anonymity), sortable away (exit/blocking, fresh accounts), or captured by platform metrics that reward performance over honesty.
The gap
The ancestral environment delivered near-total, embodied, hard-to-fake, single-identity behavioral observability inside a ~150-person reference group whose members knew one's history and held real, inescapable enforcement power. The modern environment delivers visibility that is mediated, curated, gameable, fragmented across exitable contexts, padded with non-accountable weak ties, and largely stripped of peer-enforced consequences. The reputation-tracking machinery (M11) and status-position monitoring (M5) evolved to run on dense, durable, accountable observation now operate on thin, performative, low-accountability signals, producing both reputational impunity for defection among strangers and chronic exposure of curated personas to vast unfamiliar audiences with no reciprocal knowledge of the person behind them.
Confidence
EEA-baseline side is full-text-grounded and convergent across two works in Cor's physical collection: Boehm 1999 'Hierarchy in the Forest' (ext 373 'the band keeps a dossier on every individual, noting positive and negative points'; ext 375 gossip/ridicule/ostracism as enforcement; both graded 'replicated' in Cor) and Tomasello 2014 (ext 239, bidirectional evaluation of collaborative partners; verbatim full-text quote, M11/M5). The ~150 reputational reference-group ceiling is full-text-grounded in Cor via the Dunbar works. Modern-decay side is grounded in Dunbar 2016 (Royal Society Open Science 3:150292): ext 508 (only 27.6% of Facebook friends genuine, modal 0-10%) and ext 509 (face-to-face required to prevent friendship decay through the layers), both 'replicated'. Despite the individual extractions being 'replicated'-graded, the PARAMETER is graded 'theoretical' because the specific EEA-SOC-09 claim (ancestral observability near-total/embodied/high-accountability vs modern visibility mediated/gameable/low-accountability) is an interpretive mismatch synthesis, not a single replicated measured quantity, and the 'near-total observability' baseline is qualitative. Deliberately EXCLUDED: ext 506/507 (about layer SIZES and concurrent-interaction limits, not observability) and ext 480 (QUARANTINED, snippet-sourced). The camp-shuffle/hxaro specifics and the Apicella reciprocity figures failed full-text retrieval (R2) or belong to other parameters and are NOT linked here. contested=false: reputational monitoring in small-scale societies is ethnographically well-established and uncontested; only the modern-mismatch framing is interpretive, which the 'theoretical' grade already encodes.
Evidence (5)
- ext#373 · replicated · primary
In effect, the band keeps a dossier on every individual, noting positive and negative points. Group members exercise the right to take action if a deviant begins to intimidate other group members individually - or threatens the social equilibrium of the group or its very ability to function.
- ext#375 · replicated · primary
those who have things of value but do not give are subject to social control through gossip, ridicule or ostracism.
- ext#239 · theoretical · supporting
In the hypothesis of Tomasello et al. (2012), obligate collaborative foraging became an evolutionarily stable strategy for early humans because of two interrelated processes: interdependence and social selection. The first and most basic point is that humans began a lifestyle in which individuals could not procure their daily sustenance alone but instead were interdependent with others in their foraging activities - which meant that individuals needed to develop the skills and motivations to forage collaboratively or else starve. There was thus direct and immediate selective pressure for skills and motivations for joint collaborative activity (joint intentionality). The second point is that as a natural outcome of this interdependence, individuals began to make evaluative judgments about others as potential collaborative partners: they began to be socially selective, since choosing a poor partner meant less food. Cheaters and laggards were thus selected against, and bullies lost their power to bully.
- ext#508 · theoretical · challenging
On average, respondents in Sample 1 considered that only 27.6% of their Facebook friends could be considered 'genuine' (i.e. close) friends, with a strong modal value between 0 and 10%.
- ext#509 · theoretical · challenging
it may be that face-to-face meetings are required from time to time to prevent friendships, in particular, sliding down through the network layers and eventually slipping over the edge of the 150 layer into the category of acquaintances.