Dunbar's Number¶

Cap the working set of warm, mutually-tracked social bonds an individual can hold at roughly 150, because pairwise-relationship tracking scales as O(N-squared) against a fixed neocortical budget, with nested inner shells at ~5, ~15, and ~50.

Core Idea¶

Dunbar's Number is the cognitive-neuroscience finding that the human neocortex imposes a hard upper limit of roughly 150 on the number of stable, mutually-tracking social relationships an individual can simultaneously maintain — where "stable" means warm enough that you would spontaneously intervene if the other person were in trouble and that you hold an active mental model of their identity, history, and position in your relationship network. The ceiling is not a count of contacts or acquaintances; it is a count of the simultaneous cognitive-tracking load imposed by named-individual social bonds, which scales roughly as O(N²) in pairwise-relationship information against a fixed neocortical processing budget. Robin Dunbar (1992) derived the figure from a cross-primate regression of neocortex volume against observed social group size, then found that the predicted value for humans matched modal hunter-gatherer band size, Neolithic village size before the emergence of formal coordinating roles, military company size, and measured reciprocal-interaction graphs in online social networks. Internally the 150-person layer is not flat but nested: a five-person intimate support clique, a fifteen-person sympathy group, and a fifty-person active social network are well-documented inner shells, with looser outer shells at roughly 500 (acquaintances) and 1500 (faces you can name). Crossing the 150 ceiling does not produce gradual degradation; it tends to force a qualitative regime change — coordination shifts from the implicit shared-knowledge of everyone-knows-everyone to explicit hierarchies, formal rules, and dedicated roles, because the neocortical tracking that underwrote informal coordination is at capacity and additional members cannot be held in the same active social memory.

Structural Signature¶

Sig role-phrases:

the bounded substrate — the human neocortex, a fixed processing budget for social bookkeeping that cannot be enlarged by effort
the tracked edge-type — the warm, reciprocal, named-individual bond (one you would interpose for, holding an active model of its identity and history) — not a contact or a follower
the working set — the count of those bonds held simultaneously in active social memory, distinct from the unbounded roster of acquaintances
the quadratic tracking demand — pairwise-relationship information scaling as O(N²), the load that consumes the fixed budget as the working set grows
the empirical ceiling — the measured cap near 150 where demand meets supply, derived from cross-primate neocortex-volume regression and replicated across band, village, and reciprocal-graph sizes
the nested shells — the inner (~5 clique, ~15 sympathy group, ~50 active network) and outer (~500, ~1500) layers refining the ceiling into a layered structure
the threshold regime-change — crossing the cap forces a qualitative shift, not gradual decay: implicit everyone-knows-everyone coordination gives way to explicit hierarchy, formal rules, and dedicated roles
the deliberate scope-bound — capacity not value, working set not roster, one substrate and one edge-type: it counts trackable bonds, not what any bond is worth, and does not govern software- or pheromone-coordinated systems

What It Is Not¶

Not a count of contacts, followers, or acquaintances. Eight hundred followers or a thousand named faces are real, but they are not stable bonds — only the warm, mutually-modeled tie you would interpose for carries the O(N²) tracking load the neocortical budget caps. The roster is unbounded; what the number limits is the working set of relationships actively held in social memory.
Not an exact, precise 150. The figure is a rough central estimate from a cross-primate neocortex-volume regression, with wide confidence intervals, and it sits inside a nested structure — inner shells near 5, 15, and 50 and outer shells near 500 and 1,500. Treating it as a hard, single-valued constant overstates a layered, fuzzy quantity; the layers, not the bare 150, are the real content.
Not a measure of how much relationships are worth. The ceiling bounds how many bonds can be tracked, not what any bond is worth, so it is not a social-capital metric. It caps the count over which relational value can be spread, leaving the value of each tie entirely outside its scope.
Not a target or a prescription. It is a descriptive capacity ceiling, not a recommendation that one should have ~150 relationships or a deficiency in having fewer. The figure says how many warm bonds the modal human can simultaneously track, not how many anyone ought to maintain.
Not evidence of temperament. A small active network reflects the inherited capacity of working-memory-for-people, not introversion, weak ties, or low effort — the limit is on tracking capacity, not motivation or warmth. This is why "make more friends" cannot raise the ceiling: it addresses a motivational lever where the constraint is cognitive.
Not a universal organizational threshold. The cap is bound to one substrate (the human neocortex) and one edge-type (the warm reciprocal named-individual bond); it does not govern a 10,000-person firm coordinated by software, a colony coordinated by pheromone, or a protocol coordinated by cryptography. Such systems may have their own ceiling-and-phase-transition of the general kind, but invoking "Dunbar's Number" for them borrows the vivid figure while dropping the neocortical mechanism that fixes it.

Scope of Application¶

Dunbar's Number is bounded not to "groups" in the abstract but to one substrate — the human neocortex carrying one edge-type, the warm reciprocal named-individual bond — so it lives across the human social sciences that study that substrate, while the setting changes from field to field. Its reach stays within the human case: an org coordinated by software, a colony by pheromone, or a protocol by cryptography has different bandwidth and bookkeeping, so the cap does not apply there, and that general capacity-ceiling pattern is carried by the parent prime, not by this number. The genuine habitats enumerate the in-substrate settings where the ~150 ceiling and its nested shells actually operate.

Social and cognitive psychology — the home: working-memory-for-people, the O(N²) tracking load, the nested shells (~5 support clique, ~15 sympathy group, ~50 active network), and the capacity-versus-motivation distinction that the ceiling measures.
Evolutionary anthropology — modal hunter-gatherer band size and pre-modern (Neolithic) village size before a coordinating role-layer is forced, the comparative-primate setting from which Dunbar derived the figure.
Organizational design — the small-cell move: split or cap a unit before it crosses ~150, or install a thin coordinating layer, reading a growing org's intimacy loss and coordination failures as a saturating tracking budget rather than cultural decay.
Sociology of religion — the "felt-as-a-family" congregation size before subgroup or small-group ministry becomes structurally necessary (leaf-cell splitting, megachurch small groups).
Online-community design — separating parasocial follower and contact counts from the reciprocal-interaction graph that actually carries the tracking load, and sizing moderation teams and servers-of-friends to the ceiling.
Military organization — the company-sized unit (~150) sitting between platoon and battalion, the scale at which everyone-knows-everyone coordination still holds before formal hierarchy takes over.

Clarity¶

Naming this ceiling separates two things a contact count silently fuses: the roster of people one is connected to and the working set of relationships one is actively tracking. Eight hundred followers, a thousand named faces, a wallet of business cards — these are real, but they are not 800 or 1,000 stable bonds, because only the warm, mutually-modeled tie carries the O(N²) tracking load that the neocortical budget caps. Once a practitioner holds that distinction, the question "how large is your network?" splits into a layered one — how many in the support clique (~5), the sympathy group (~15), the active network (~50), the band (~150), the acquaintance shell (~500) — and the figures stop competing. A claim of vast reach and a claim of capacity-limited intimacy are no longer contradictory; they are statements about different shells.

The number also separates capacity from motivation and from value. A small active network is not evidence of an introverted temperament or of weak ties; it is a property of working-memory-for-people that the modal human inherits regardless of effort or warmth, so "make more friends" cannot move the ceiling the way it could move a motivational deficit. And because the limit governs how many bonds can be tracked, not how much any bond is worth, it does not measure social capital — it bounds the count over which capital can be spread. The sharpest question Dunbar's Number licenses is structural rather than personal: as a group approaches 150, is its coordination still running on everyone-knows-everyone shared knowledge, or has the tracking budget already saturated — so that the felt loss of intimacy and the proliferation of "I didn't know we had a team for that" are not cultural decay but the signature of a capacity limit about to force formal roles, hierarchy, or a subdivision?

Manages Complexity¶

A group's social and coordinative life is, at face value, an unbounded mess of particulars — who talks to whom, which subgroups feel like family and which feel like strangers, why a thirty-person team runs on tacit understanding while a three-hundred-person one drowns in process, when a congregation or a band or a startup tips from informal warmth into formal structure, why "make more friends" advice fails for some people and lands for others. Each case looks like it needs its own social, cultural, or motivational explanation. Dunbar's Number collapses that sprawl onto a single governing quantity: the count of warm, mutually-modeled, named-individual bonds an individual is actively tracking against a fixed neocortical budget, with a known ceiling near 150 and known inner shells (~5, ~15, ~50) and outer shells (~500, ~1500). Instead of re-deriving each social configuration from scratch, the analyst tracks one number — where the active-bond count sits relative to those layer boundaries — and reads the qualitative outcome off it.

The compression is sharp because the ceiling is a threshold, not a slope, so it imposes a small branch structure rather than a continuum of cases. Below 150 the analyst predicts that coordination can still run on everyone-knows-everyone shared knowledge, intimacy is felt, and no dedicated coordinating role is structurally forced. As the active-bond count approaches 150 the tracking budget saturates, and the analyst predicts the regime change: implicit coordination gives way to explicit hierarchy, formal rules, and dedicated roles, with the felt loss of intimacy and the "I didn't know we had a team for that" failures appearing not as cultural decay but as the signature of the capacity limit biting. The inner shells refine the same read at smaller scale — a support clique of five, a sympathy group of fifteen, an active network of fifty — so a question as open-ended as "how is this person's or this group's social world structured?" reduces to locating one figure among a handful of boundaries and reading off which coordination regime, which depth of intimacy, and which design pressure (subdivide, formalize, or stay flat) obtain.

Two further confusions the number dissolves are themselves complexity reductions: it separates the roster (contacts, followers, named faces — unbounded) from the working set (tracked bonds — capped), so that claims of vast reach and capacity-limited intimacy stop competing and become statements about different shells; and it separates capacity from motivation and value, so the analyst does not need a temperament theory or a social-capital audit to explain a small active network — it is a property of working-memory-for-people the modal human inherits. What would otherwise demand a bespoke account of each person's psychology or each group's culture becomes a reading off one layered count against a fixed budget.

Abstract Reasoning¶

Dunbar's Number licenses reasoning that treats a group's social and coordinative life as governed by one count — the warm, mutually-modeled, named-individual bonds being actively tracked against a fixed neocortical budget — read against a ceiling near 150 and a set of nested inner (~5, ~15, ~50) and outer (~500, ~1500) shells. The analyst reasons from where that count sits to the coordination regime, and back from coordination symptoms to whether the capacity limit is biting.

Diagnostic (read a regime change as a capacity limit, not cultural decay; place a network among shells). The defining inference re-reads organizational and social symptoms through the ceiling. When a growing group shows a felt loss of intimacy, balkanizing channels, and "I didn't know we had a team for that" failures, the analyst diagnoses a tracking budget that has saturated near 150 — the neocortical bookkeeping that underwrote informal coordination is at capacity — rather than a failure of culture or goodwill. A small active network is diagnosed as a property of working-memory-for-people, not as introverted temperament or weak ties, because the limit is on tracking capacity, not motivation or warmth. And a person's social world is located by shell: a five-person support clique, a fifteen-person sympathy group, a fifty-person active network, a 150-person band, a 500-person acquaintance layer — so a claim of vast reach and a claim of capacity-limited intimacy are diagnosed as statements about different shells rather than as contradictions. The inference runs coordination breakdown → tracking budget saturated (and small network → inherited capacity), never breakdown → "the culture decayed" or small network → "weak character."

Interventionist (engineer the split or formalization before the ceiling bites). Because the ceiling is a threshold a group crosses, the design lever is to anticipate the regime change rather than absorb it. A unit approaching 150 is predicted to be forced from everyone-knows-everyone shared knowledge into explicit hierarchy, formal rules, and dedicated roles; so the intervention is to pre-engineer the split into smaller cells, or to install a thin coordinating layer, before the breakdown manifests. Conversely, holding a unit deliberately below the ceiling is predicted to preserve implicit coordination and felt intimacy. The lever the concept rules out is equally informative: "make more friends" cannot raise the active-bond ceiling, because it addresses motivation where the constraint is capacity. Each design move pairs a manipulation of unit size (split, cap, or layer) with a predicted coordination regime.

Boundary-drawing (working set, not roster; capacity, not value; a threshold, not a slope; one substrate). The concept fixes its scope through several lines. It is a count of the working set of actively tracked bonds, not the roster of contacts, followers, or named faces — only the warm, mutually-modeled tie carries the O(N²) tracking load that the budget caps, so a wallet of business cards is outside the regime. It bounds capacity, not value: it limits how many bonds can be tracked, not what any bond is worth, so it does not measure social capital. It behaves as a threshold, not a continuum — below 150 no coordinating role is structurally forced; at the ceiling the regime changes qualitatively — which is why the analyst reasons in branches rather than along a smooth curve. And it is bound to one substrate and one edge-type (the human neocortex; the warm reciprocal named-individual bond), so it does not govern an organization coordinated by software or a colony coordinated by pheromone, where the bandwidth and bookkeeping differ.

Predictive / branch-ordering. From where the active-bond count sits relative to the layer boundaries the analyst forecasts the qualitative state before investigating: below 150, implicit coordination, felt intimacy, no forced role-layer; approaching 150, saturation and the predicted shift to hierarchy and formal process; within the inner shells, the depth of intimacy available at five, fifteen, and fifty. Coordination regime, intimacy depth, and the design pressure (subdivide, formalize, or stay flat) all read off one layered count against a fixed budget.

Knowledge Transfer¶

Within psychology and the human social sciences the finding transfers as mechanism, but it is important to see what the home domain is: not "groups" in the abstract but the single human-neocortex substrate carrying one edge-type — the warm, reciprocal, named-individual bond. Across the subfields that study that substrate the layered ceiling carries intact, with its diagnostics and interventions unchanged. In anthropology it predicts modal hunter-gatherer band size and pre-modern village size before a coordinating role-layer is forced. In organizational design it licenses the small-cell move — split a unit before it crosses ~150, install a thin coordinating layer — and reads a growing org's intimacy loss and "I didn't know we had a team for that" failures as a saturating tracking budget rather than cultural decay. In the sociology of religion it locates the felt-as-a-family congregation size before subgroup ministry becomes necessary; in online-community design it separates parasocial follower counts from the reciprocal-interaction graph that actually carries the O(N²) load. Across all of these the substrate is constant; only the setting changes, so the number, the shells (~5, ~15, ~50, ~150, ~500, ~1500), and the prediction of a qualitative regime change at the ceiling all carry without translation. This is genuine mechanism transfer, and it is the proper reach of the concept.

Beyond that substrate the honest reading is largely shared abstract mechanism, not the named concept — and the entry itself is unusually explicit about this. The general pattern that recurs across domains is the one the entry instantiates: a capacity-bounded substrate holds a working set of named entities with quadratic pairwise-tracking demand against a fixed budget, yielding a sharp ceiling whose crossing forces a regime change (working memory for chess positions, attention across channels, maintenance of distinct conversational threads all share this shape). That general pattern travels because it is substrate-neutral. What does not travel is the home-bound cargo that makes the concept Dunbar's Number specifically: the figure 150, the neocortex-volume regression across primates that derives it, the particular nested shells, and the warm-reciprocal-bond edge-type. An organization of 10,000 employees coordinated by software, an ant colony coordinated by pheromone, a protocol coordinated by cryptography are not in the same regime — the substrate has different bandwidth and different bookkeeping, so the cap does not apply, even though each may have its own ceiling-and-phase-transition of the general kind. The correct cross-domain lesson therefore carries the parent pattern (bounded tracking capacity → ceiling → formalization/subdivision), not the number; reaching for "that team has hit its Dunbar's Number" outside the human case is at best a pointer to the general pattern and at worst (A) loose analogy that borrows the vivid figure while dropping the neocortical mechanism that fixes it. The discipline is to invoke the general capacity-ceiling pattern where the substrate differs, and to reserve "Dunbar's Number" for the one substrate whose budget it actually measures (see Structural Core vs. Domain Accent).

Examples¶

Canonical¶

Robin Dunbar's 1992 derivation is the defining construction. Working in comparative primatology, he plotted the neocortex ratio (neocortex volume relative to the rest of the brain) against mean observed social-group size across a range of primate species and found a tight positive regression: species with proportionally larger neocortices live in larger cohesive groups. Reading the human neocortex ratio off that regression line yields a predicted group size of roughly 150. Dunbar then checked the number against independent human data and found convergence — modal hunter-gatherer clan sizes, the size of Neolithic farming villages, and the personnel of a military company all cluster near it. The figure is thus not a headcount but the value at which a species' fixed neocortical bookkeeping budget is predicted to run out for tracking mutually-known individuals.

Mapped back: The neocortex is the bounded substrate; the mutually-known group member is the tracked edge-type; and the regression's output, ~150, is the empirical ceiling — the point where the fixed budget meets the rising tracking demand. That the demand rises steeply (pairwise relationships) rather than linearly is the quadratic tracking demand, and reading the number off a regression rather than treating it as an exact constant reflects the nested shells and fuzzy layered structure the derivation actually supports.

Applied / In Practice¶

The number is a working design rule among Hutterite agricultural colonies, a case Dunbar himself cites. These communal farming settlements have a long-standing practice of splitting a colony in two once it grows to around 150 members. Their stated rationale is strikingly close to the mechanism: below that size the colony can be governed by mutual obligation and everyone-knows-everyone social pressure, but beyond it, they find, informal control breaks down and the community would need formal police-like enforcement and hierarchy to hold together — so they pre-empt the transition by dividing while still cohesive. The practice long predates Dunbar's theory, which is what makes it corroborating rather than derived.

Mapped back: The colony's members bound by mutual obligation are the tracked edge-type on the human bounded substrate; ~150 is the empirical ceiling. The Hutterites' observation that informal control fails past that size and would require formal enforcement is precisely the threshold regime-change — a qualitative shift from everyone-knows-everyone coordination to explicit hierarchy — and their deliberate pre-emptive split is the interventionist "engineer the split before the ceiling bites" move the concept licenses.

Structural Tensions¶

T1: Sharp threshold versus fuzzy layered estimate (the crisp number and its wide error bars). The concept's usefulness comes from a crisp ceiling — 150, a threshold that forces a qualitative regime change rather than gradual decay — which lets an analyst read coordination state off one boundary. But the same entry insists the figure is a rough central estimate from a cross-primate regression with wide confidence intervals, sitting inside a nested, fuzzy structure whose layers (~5, ~15, ~50, ~500, ~1500) are the real content. These pull against each other: the more one leans on 150 as a hard line to make predictions, the more one overstates a quantity the derivation supports only as a fuzzy layered range; the more one honors the fuzziness, the less the "threshold" does the sharp diagnostic work claimed for it. Diagnostic: Is the analysis using 150 as a precise cutoff (overstating a fuzzy estimate), or as one boundary in a layered range whose neighbors matter as much as the central figure?

T2: Fixed neocortical budget versus institutional augmentation (a hard limit humans routinely exceed). The mechanism rests on a budget that cannot be enlarged by effort — "make more friends" cannot move it. Yet humans coordinate in firms, cities, and armies of millions, precisely by offloading the tracking the neocortex cannot hold onto external scaffolding: writing, org charts, contact lists, roles, reputation systems. The tension is that the limit is real for warm mutually-modeled bonds held in active social memory but is routinely and massively transcended for coordination by substituting formal bookkeeping for neocortical bookkeeping. So the ceiling both binds (intimate active networks stay small) and does not bind (organizations scale indefinitely) depending on which is being counted — and the interesting cases are exactly where augmentation partly substitutes for the bond, blurring whether the budget is really the constraint. Diagnostic: Is the group relying on warm mutually-tracked bonds (the budget binds near 150) or on external scaffolding — roles, records, reputation — that substitutes for that tracking (the ceiling is transcended, not violated)?

T3: Descriptive ceiling versus prescriptive design rule (an average pressed into a mandate). The number is explicitly not a target — it describes how many bonds the modal human can track, not how many anyone ought to maintain, and having fewer is no deficiency. Yet its most active use is prescriptive: split a unit at 150, cap a colony, size a congregation. That crossing from "is" to "ought" is where the concept earns its organizational keep, but it is also where it is most abused — reifying a fuzzy population average into a hard managerial threshold, ignoring the wide individual variation the average conceals, and treating a single number as a design constant. The tension is that the descriptive honesty ("rough, layered, individual-varying") and the prescriptive utility ("split here") sit uneasily on the same figure. Diagnostic: Is 150 being used to explain an observed coordination shift (descriptive, its proper role) or to mandate a unit size as if the average bound every individual (prescriptive over-reach)?

T4: Working set versus roster (a clean distinction that resists measurement). The concept's first move is to split the unbounded roster (contacts, followers, named faces) from the capped working set (warm, mutually-modeled, would-interpose-for bonds), and only the latter carries the O(N²) load. Conceptually this dissolves the apparent contradiction between vast reach and small intimacy. But in practice the boundary between a tracked bond and a mere acquaintance is a gradient, not a line — ties warm and cool, get modeled more or less richly, shade continuously from the sympathy group out to the acquaintance shell. So the very quantity the ceiling caps is the one hardest to count, and any empirical test must operationalize "warm mutually-tracked" with a proxy that inevitably admits some roster and excludes some genuine bonds. The clean distinction that makes the number coherent is the thing that makes it hard to measure. Diagnostic: For the ties being counted, is there an operational test of mutual, warm, actively-modeled tracking — or is the count quietly sliding back toward the unbounded roster it was meant to exclude?

T5: Saturation diagnosis versus alternative causes (reading every breakdown through the ceiling). The signature diagnostic re-reads intimacy loss, balkanizing channels, and "I didn't know we had a team for that" as a tracking budget saturating near 150 — not cultural decay, not bad management. This is powerful because it converts moralized complaints into a structural read. But it is also a lens that can over-explain: coordination can break down from genuine mismanagement, misaligned incentives, or toxic culture at any size, and attributing every large-group friction to N crossing 150 risks excusing fixable failures as inevitable capacity limits. The tension is that the same re-reading which rescues a group from false blame ("your culture didn't decay") can also absolve real, addressable dysfunction by naming it a law of the neocortex. Diagnostic: Is the coordination failure tracking group size crossing the ceiling (capacity biting), or would it persist at half the size — pointing to incentives, leadership, or culture the size-story would wrongly excuse?

T6: Autonomy versus reduction (the named number or the general capacity-ceiling pattern). "Dunbar's Number" is a specific, canonically derived finding whose irreducible cargo is the figure 150, the cross-primate neocortex-volume regression that fixes it, the particular nested shells, and the warm-reciprocal-bond edge-type — all bound to one substrate, the human neocortex. In situ that specificity is exactly the point. But what actually travels across domains is the parent pattern it instantiates: a capacity-bounded substrate holding a working set of named entities with quadratic pairwise-tracking demand against a fixed budget, yielding a ceiling whose crossing forces a regime change — a pattern equally present in working memory, attention across channels, and conversational threads, none of which have a "150." A software-coordinated firm or a pheromone-coordinated colony may have its own ceiling-and-phase-transition, but it is not Dunbar's Number. The tension is between a vivid named constant that earns its own study and the recognition that its portable insight is the substrate-neutral pattern, not the figure. Diagnostic: Resolve toward the general capacity-ceiling pattern (bounded tracking → ceiling → formalization) when the substrate differs from the human neocortex; toward named Dunbar's Number when counting warm reciprocal bonds in the one substrate whose budget it measures.

Structural–Framed Character¶

Dunbar's Number sits toward the structural end of the structural–framed spectrum but stops short of the pole — best read as mixed-structural: a genuine capacity-ceiling mechanism wearing heavy cognitive-neuroscience vocabulary. On the five criteria its structural credentials are strong. Its evaluative weight is nil — the entry is emphatic that the figure is a descriptive capacity ceiling, not a target or a prescription, and that a small active network is no deficiency and no evidence of temperament; ~150 describes how many warm bonds the modal human can track, praising and blaming nothing. It is not human-practice-bound — remove every psychologist and the neocortical budget still caps warm mutually-modeled bonds near 150, the nested shells still hold at ~5, ~15, ~50, and groups still hit the regime change at the ceiling; the Hutterite colonies that split at 150 did so long before Dunbar's theory, which is exactly what makes their practice corroborating rather than derived. Its institutional origin is none: the ceiling is a fact of how a fixed neocortical bookkeeping budget meets an O(N²) tracking demand, read off a cross-primate neocortex-volume regression — Dunbar measured a limit the brain already imposes, he did not legislate it. And within its proper range — the human social sciences that study the one neocortical substrate — cross-domain reuse is recognition rather than import: moving from anthropology (band size) to organizational design (the small-cell split) to the sociology of religion (congregation size) to online-community design, the same ceiling and its shells are recognized intact, with only the setting changing. These four marks place it firmly on the structural side, closely analogous to how isostasy is characterized — a real, evaluatively neutral, recognized-in-nature limit.

What keeps it off the structural pole is the remaining criterion, vocab_travels, which it fails, together with the fact that it is pinned to a single substrate. The operative content is irreducibly tied to the human neocortex — the neocortex ratio, the cross-primate regression that fixes the figure, the O(N²) pairwise-tracking load, the particular value 150, the specific nested shells (~5, ~15, ~50, ~500, ~1500), the warm reciprocal named-individual edge-type — and none of it floats free of that substrate. Ported outward, "that team has hit its Dunbar's Number" keeps only the bare bounded-tracking-and-ceiling shape and renames every component: a 10,000-person firm coordinated by software or a colony coordinated by pheromone has different bandwidth and bookkeeping, so the cap simply does not apply, and invoking the vivid figure there is analogy, not mechanism. The portable structural skeleton it shares — a capacity-bounded substrate holds a working set of named entities with quadratic pairwise-tracking demand against a fixed budget, yielding a sharp ceiling whose crossing forces a regime change — is genuinely substrate-independent (working memory, attention across channels, and conversational threads all show it), but it is exactly the part the catalog already carries as the general capacity-ceiling / threshold (regime-change) pattern that Dunbar's Number instantiates. What is distinctive to "Dunbar's Number" — the figure 150, the neocortex-volume derivation, the particular shells, the warm-bond edge-type — is the domain-accented expression that does not travel. Its character: structural in skeleton — a real, evaluatively neutral, recognized-in-nature capacity ceiling whose crossing forces a coordination regime change — but stated in cognitive-neuroscience vocabulary and fixed to the human-neocortex substrate, leaving it mixed-structural rather than a free-floating prime.

Structural Core vs. Domain Accent¶

This section settles why Dunbar's Number is a domain-specific abstraction rather than a prime, by peeling the substrate-neutral capacity-ceiling structure away from the neocortical measurements that fix a particular figure to it.

What is skeletal (could lift toward a cross-domain prime). Strip the brain away and a thin relational structure survives: a fixed-budget substrate holds a working set of named entities whose mutual tracking scales super-linearly with the set size, so demand overtakes supply at a determinate boundary, and crossing that boundary forces a qualitative regime change rather than a graceful decline. The portable pieces are abstract: a fixed processing budget, a working set distinct from an unbounded roster, a quadratic (pairwise) load that consumes the budget as the set grows, and a threshold whose crossing flips the coordination regime. That skeleton is genuinely substrate-portable — it recurs in working memory for chess positions, attention across channels, and the maintenance of distinct conversational threads — which is exactly why the catalog already carries it as the general capacity-ceiling / threshold (regime-change) pattern that Dunbar's Number instantiates. This is the core the finding shares, not what makes it distinctive.

What is domain-bound. Almost all of the operative content is cognitive-neuroscience furniture, and none of it survives extraction intact: the human neocortex as the specific bounded substrate; the neocortex ratio and the cross-primate regression of neocortex volume against group size that derives the figure; the particular value ~150; the specific nested shells at ~5, ~15, ~50, ~500, and ~1,500; and the load-bearing edge-type — the warm, reciprocal, named-individual bond you would interpose for, held in active social memory. The decisive test: remove the neocortical budget and the warm-bond edge-type and it is no longer Dunbar's Number but a bare threshold — a 10,000-person firm coordinated by software or a colony coordinated by pheromone has different bandwidth and bookkeeping, so the cap simply does not apply, and invoking "150" there borrows the vivid figure while dropping the mechanism that sets it. The regression, the figure, and the shells are all specific to the one human substrate whose budget the number measures.

Why this does not clear the prime bar. A prime is a relational structure whose vocabulary travels and whose cross-domain transfer is recognition of the same mechanism, not analogy. Dunbar's Number's transfer is bimodal — and, like other single-substrate findings, its home is one substrate: the human neocortex. Within that substrate the mechanism travels intact across every field that studies it — anthropology (band size), organizational design (the small-cell split), the sociology of religion (congregation size), online-community design — where the number, the shells, and the regime-change prediction port without translation because only the setting changes; that is recognition of the same limit. Beyond the neocortex it travels only by renaming components and dropping the neocortical budget that fixes the figure — a software- or pheromone-coordinated system may have its own ceiling, but it is not this one, so reaching for "that team hit its Dunbar's Number" is analogy, the boundary between the two. And when the bare structural lesson is needed cross-domain — a bounded tracking budget yields a ceiling whose crossing forces formalization or subdivision — it is already supplied, in more general form, by the parent it instantiates: the capacity-ceiling / threshold (regime-change) pattern, which carries across working memory, attention, and conversational threads without any "150." The cross-domain reach belongs to that parent; "Dunbar's Number," as named, carries neocortical baggage — the figure, the primate regression, the particular shells, the warm-bond edge-type — that does not and should not travel past the human case.

Relationships to Other Abstractions¶

Current abstraction Dunbar's Number Domain-specific

Parents (1) — more general patterns this builds on

Dunbar's Number is a kind of Constraint Prime

Dunbar's number specializes constraint to a cognitively bounded working set of mutually tracked social relationships.

Hierarchy path (1) — routes to 1 parentless root

Dunbar's Number → Constraint

Not to Be Confused With¶

Miller's "magical number seven" / working-memory span. The classic finding that short-term memory holds roughly 7 (±2) chunks at once. Both are cognitive-capacity limits stated as a number, but they measure different systems: Miller's is momentary holding capacity for arbitrary items in working memory, whereas Dunbar's is long-term tracking capacity for warm, mutually-modeled social bonds. Tell: is the limit on how many items you can juggle right now (Miller), or on how many named individuals you can hold an active social model of over time (Dunbar)?
Social capital. The value or resources embedded in one's social ties — the benefits that flow from a network's structure and the trust within it. Dunbar's Number bounds how many bonds can be tracked, not what any bond is worth, so it caps the count over which social capital is spread without measuring that capital. Tell: are you asking how much a person's relationships are worth to them (social capital), or how many warm bonds they can simultaneously hold in mind (Dunbar)?
Span of control. The management-theory limit on how many direct reports one supervisor can effectively oversee (often cited as ~5–10). It concerns authority/supervisory relationships within a formal hierarchy; Dunbar's Number concerns warm, reciprocal peer bonds and predicts when informal coordination gives way to hierarchy at all. Span of control operates inside the formal structure that crossing Dunbar's ceiling forces into being. Tell: is the count of subordinates a boss can manage (span of control) or of peers one can warmly track (Dunbar)?
Six degrees of separation / small-world reach. The network property that any two people are connected by a short chain of acquaintances — a fact about global path length and reach across a whole network. Dunbar's Number is about the size of one individual's own actively-tracked set, not how few hops separate distant strangers. A person embedded in a small-world network still tracks only ~150 warm bonds. Tell: is the claim about how far the network reaches through weak-tie chains (small-world), or how many bonds one node can hold (Dunbar)?
Capacity-ceiling / threshold (the parent it instantiates). The substrate-neutral skeleton — a fixed-budget substrate holding a working set with super-linear tracking demand, yielding a ceiling whose crossing forces a qualitative regime change — belongs to this parent (with threshold for the regime-change), which carries the lesson to working memory, attention across channels, and conversational threads, none of which have a "150." Dunbar's Number is the human-neocortex specialization. Tell: when the substrate is software-, pheromone-, or cryptography-coordinated rather than the human neocortex, the portable structure is the capacity-ceiling/threshold pattern; "Dunbar's Number" there is analogy borrowing the vivid figure. (Treated fully in earlier sections.)

Neighborhood in Abstraction Space¶

Dunbar's Number sits in a moderately populated region (53^rd percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.

Family — Social Perception & Self-Referential Bias (23 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-07-12