← all systems · see it in The System
Friendship
Friends are pairs of individuals that engage in bi-directional affiliative (nonaggressive, nonreproductive) interactions with such frequency and consistency so as to differentiate them from nonfriends; in people, a relationship of mutual affection.
Properties published scienceour proposal
Sourced fields, and our own readings.
| asymmetry | about half of named ties are not named back (Almaatouq) |
| born at | per pair: weeks after meeting; for the class: contested (C-3740) |
| born precision | weeks (per pair); unknown (class) |
| boundary | Inside: the tie between two individuals who are not (or not only) kin or mates: how they meet, recognise and choose each other, the history of exchanges that builds the bond, what holds it (loose reciprocity, trust, stress buffering), what it means to them and their culture, and what the pair does together. Outside: the two persons themselves (sys:human, or each animal), the groups and networks the tie sits in (peer group, social network), kinship and pair bonds, and the culture that sets how freely friends are chosen; each walked or queued separately. Peer acceptance, popularity and status in the group belong to the peer-group walk; mutual dyadic ties to this one (a scope line, not a sourced claim). |
| cheater | short-changing in continuous exchanges; imbalances tolerated up to a point (Brent; Carter) |
| drive | driven |
| enforcement | partner-choice |
| flips when | a rival, a shock such as discovering incompatible values, or a better outside option (switching) |
| joint output | a held bond that gives each tolerance, help, support in conflicts and stress buffering, and in animals more surviving offspring |
| knockout | Disruptions of friendships are associated with increased guilt, anger, and depression, and may be highly stressful events, especially in childhood. |
| nested level | A dyadic social system: built from two partners and their exchanges; one tie of a social network and the unit of a peer group's cliques; runs on its members' acts and talk (Human L7 in people). |
| obligate | facultative |
| partners | sys:human, sys:human |
| recurrence | repeated exchanges over years; persists across distance for 8 years or more |
| regime | starts with testing the waters and hours together; ends by drift, shock or switching |
| saturation | about five close friends in children (Shen); preferred layers 3-5, 9-15 (Zhou) |
| scale | social |
| separability | two individuals with their own lives, lineages and other ties |
Measures (22) published science
| survival odds with stronger social relationships | 1.50 (95% CI 1.42-1.59); 1.91 for complex measures of social integration odds ratio |
| reciprocity of named friendships | about half; 53% (413 of 775) in one study, 34-53% across five datasets; 64% among adolescents naming up to 10 best friends share of named ties named back |
| time to form a friendship | about 20-60 hours in the first weeks; over 100 hours for a close friend |
| individual recognition memory, dolphins | up to 20 years |
| friendship persistence across distance | 8 or more years |
| dolphin bond tracking | 59 males followed 14 years, adolescence (8-14) to adulthood (15-21) |
| relationship formation, monk parakeets | 22 birds from four groups followed over 22 days |
| homophily onset after a transition | self-segregation within one term for two-thirds of 16 traits; cohorts of about 120 students terms |
| genetic similarity of friends | at the level of fourth cousins relative to strangers kinship coefficient equivalent |
| close friends beyond which benefits stop (children) | about 5 close friends |
| preferred group sizes (layers) | 3-5, 9-15, 30-45 (scaling ratio near 3) people |
| Dunbar's number | 150 proposed (range 100-250); one re-analysis gave 95% intervals of 4-520 and 2-336 stable relationships |
| children with at least one friend | 75% of preschoolers; 78% through fifth grade (co-nomination); 55% with a mutual best friend; about 15% chronically friendless share of children |
| adults reporting no close friend (US, 2023) | 8% none; a further 7% only one share of respondents |
| relational mobility survey | 39 societies |
| neural homophily cohort | 279 graduate students, 100% response people |
| adolescent friendship networks (Add Health) | 13,465 adolescents |
| adult network trajectory | 184 participants aged 18-94 at start, followed 10 years people |
| characteristic clock: hours spent with new acquaintances | first-year students reported three times over 9 weeks |
| time to form a friendship (Hall 2018) | casual friend: about 40-60 hours in the first 6 weeks (43 h in Study 2; 94 h in Study 1); friend: about 80-100 hours (57 and 164 h in the two studies); good or best friend: 119 hours over 3 weeks, 219 over 3 months |
| friendship stability (children and adolescents) | .48 of friendships stable between measurement occasions proportion |
| support clique and sympathy group | support clique typically 3-5; sympathy group 12-20, contacted at least monthly people |
Open contentions (8) open contention
Where sources or readings disagree. Both stay on the map.
- C-394 Do behaviours spread through social networks, or do similar people just cluster together?
booked: The map keeps both: evidence of real spread from trials, and the point that ordinary observations confuse spread with similarity.
the other reading: homophily and contagion are generically confounded in observational data (Shalizi & Thomas 2011); separating them needs strong assumptions (SIENA models, Kempnich 2026) - C-413 Where do thoughts come from?
booked: No theory is proven, so the map lists the main ones side by side and doesn't pick one. It shows where thinking is handled in the brain, not where thoughts start.
the other reading: see the registry - C-3615 Who shapes a teenager more: one close friend or the wider group?
booked: The map keeps the competing readings; none is ruled.
the other reading: see the registry; the peer-group walk holds both - C-3736 Why are friends genetically alike: do people choose similar friends, or do schools and neighbourhoods sort them first?
booked: The map keeps the competing readings; none is ruled.
the other reading: about half of friends' genetic similarity is shared by schoolmates generally: social structuring by school assignment is a principal source, beside social homophily (Domingue 2018) - C-3737 What counts as a friendship: can kin and partners be friends, and do animals have friends?
booked: The map keeps the competing readings; none is ruled.
the other reading: ties to individuals who are neither kin nor mates (Christakis & Fowler; the roster's line); friendship and romance are not that different and pair bonds may have begun in friendship (Sandel 2025); 'friendship' for animal bonds risks anthropomorphism (Sandel); developmental psychology: friendships are 'uniquely voluntary dyadic relationships of affection', operationalised as mutual nomination (Meter & Card 2016, citing Furman); Silk 2002: reserve the word for non-kin bonds meeting set criteria; Seyfarth & Cheney 2012 (abstract): applying it to animals is not anthropomorphic - C-3738 What is friendship for?
booked: The map keeps the competing readings; none is ruled.
the other reading: each reading above is the alt of the others - C-3739 Do friends make people live longer, or do healthy people have more friends?
booked: The map keeps the competing readings; none is ruled.
the other reading: cause and effect not established; candidate pathways listed (healthier lifestyles, help-seeking, coping, physiology) (encyclopedia) - C-3747 Do friends start out alike, or does likeness matter less as the years go by?
booked: The map keeps the competing readings; none is ruled.
the other reading: homophily fades over years and later friendships rest on propinquity (Godley 2008, reported by Kempnich)
Cross-scale patterns seen here (18) our proposal
Shapes that recur when the same kind of problem is solved at different scales, and how each shows in Friendship. Each links to every other system that shows it.
- not-diluting wins (structural)
With age, peripheral partners thin out while close partners hold steady, so close friends come to make up a greater share of the network. - same role, different builds (structural)
One job, a lasting preferential tie to a non-mate that brings help, is built from grooming in primates, regurgitated blood in vampire bats, whistles and alliances in dolphins, proximity in mares and talk in people. - the whole sets the local reference (structural)
The society sets the reference each pair is read against: relational mobility fixes how freely ties are chosen, and who counts as a friend differs (a Russian's one or two friends against a Canadian's many). - trade one property for another (structural)
Breadth bought with depth: bats balance new partners against focused investment (social bet-hedging); men may have more friends but less time with each; past about five close friends children's cognition was lower. - Umwelt: each system carves its own slice of one shared medium (structural)
Each species carves the bond from its own senses: dolphins from whistles remembered for 20 years, primates from faces and grooming, bats from shared blood, people from talk. - who opens the door (structural)
The chooser's preference for similar others is in place before the newcomer arrives (selection for homophily at the outset), and a stranger's no-contact approach waits on the other's tolerance: the receiver's side opens the door. - borrowed-then-held (partial)
Friends first supplied by mothers or parents (young macaques play with the offspring of their mothers' friends; children inherit their parents' friends), later chosen by the person on similarity and values. - coordination-before-mapping (partial)
Exchanges between friends kept in balance by emotional bookkeeping, an attitude toward the partner, rather than a tally of past moves. - nested ladder (partial)
Human groups come in a discrete hierarchy of preferred sizes (3-5, 9-15, 30-45), each layer about three times the last. - same shape, opposite sign (partial)
One process, friend selection acting on genotype, runs with sign reversed by gene set: friends match on olfactory genes and differ on immune genes. - symmetry breaking (name the cut) (partial)
Two people's ties look symmetric but about half run one way; the direction decides whether influence passes. - templating (partial)
The oxytocin system that serves mother-infant and pair bonds is reused for friendships; a shared suite of circuits works across humans and other animals. - The carry (partial)
The wait: a friendship forms only after enough hours together (about 20-60, over 100 for a close one); strangers move up from no-contact to contact. - causal ceiling (surface)
People cannot perceive whether their friendships are returned from inside the tie; the network reveals it only from outside. - lightning (surface)
A friendship can end in a sudden shock, such as learning that a friend holds incompatible values, after which the bond is not the same. - namespace completeness (the map predicts its gaps) (surface)
The layer series (3-5, 9-15, 30-45, etc.) predicts the next sizes from its ratio. - partner-becomes-part (surface)
Unrelated friends treated as 'functional kin'; genetically as alike as fourth cousins; but they keep separate inheritance. - same toolkit, independent builds (surface)
Bonds in birds and mammals run on a shared suite of circuits and neuromodulators; the suite is read as homologous, so independence is not shown.
Valence: what it trades, and with whom (17) published science
Each exchange with a neighbour, read from both sides: who gains, who pays, and what carries it.
- Oxytocin signalling via sys:sensory-environment · Friendship gain
- Human (Homo sapiens) via sys:sensory-environment · Friendship gain
- Human brain via sys:sensory-environment · Friendship gain
- Social network via sys:sensory-environment · Friendship gain
- Vampire bat via sys:sensory-environment · Friendship gain
- Chimpanzee (Pan troglodytes) via sys:sensory-environment · Friendship gain
- Endogenous opioid system via sys:sensory-environment · Friendship gain
- Horse (reintroduced) via sys:sensory-environment · Friendship gain
- HPA axis via sys:sensory-environment · Friendship gain
- Culture via sys:sensory-environment · Friendship mixed
- Bottlenose dolphin (Tursiops) via sys:sensory-environment · Friendship gain
- Peer group via sys:sensory-environment · Friendship gain
- Attachment (the infant-caregiver bond) via sys:sensory-environment · Friendship gain
- Kinship via sys:sensory-environment · Friendship mixed
- Baboon via sys:sensory-environment · Friendship gain
- Pair bond via sys:sensory-environment · Friendship mixed
- Monk parakeet via sys:sensory-environment · Friendship gain
Partners and parts published science
Built from / builds (5)
- Human (Homo sapiens)
- Similarity between the two
- Exchanges of help and grooming
- Peer group
- Two people competing for one friend's closeness (friendship jealousy)
Sits inside / holds (3)
expressed by (11)
- Symbiosis/Co-Creation (Collaborative Specialization)
- Sensing/Scanning
- Uncertainty Reduction/Learning
- Risk-managed Exploration
- Optimal Stopping & Patch-Switching
- Social Learning / Public Information Use
- Divide & Isolate
- Validation of Identity/Intention
- Redundancy & Diversity
- Triage & Prioritized Allocation
- Bridging Supply Gaps (Store & Dormancy)
floor filled by (14)
- Society as sociologists see it (sociology's object: social structure and social action)
- Oxytocin signalling
- Human (Homo sapiens)
- Language (human language)
- Human brain
- School
- Social network
- Chimpanzee (Pan troglodytes)
- Endogenous opioid system
- HPA axis
- Culture
- Bottlenose dolphin (Tursiops)
- Peer group
- Similarity between the two
Kinds (8)
- Peer group
- Friendship of pleasure
- Friendship of advantage
- Friendship of virtue
- Cross-sex friendship
- Best friendship
- Social bond in other animals
- Male dolphin alliance
wired (2)
Notes published science
- Only about half the friendships people name are named back. Across five surveys, between 34% and 53% of named friends returned the nomination, though people expect almost all of theirs to be mutual. And a friend's nudge to change a habit worked only when the tie ran from the nudger to the person nudged, or both ways (src:almaatouq-2016).
- Friends are genetically about as alike as fourth cousins, though they are not related, and they tend to match on a set of smell genes and differ on a set of immune genes (src:christakis-fowler-2014). But a later study of about 5,500 American teenagers found that about half of that likeness is shared by schoolmates in general: who ends up in the same school sorts people before they choose their friends (src:domingue-2018).
- In wild chimpanzees, grooming raised oxytocin only when the partner was a bonded friend, related or not: in the review's words, what mattered 'was not grooming in general, but grooming with a friend' (src:brent-2014).
- Hours make friends, but talk decides which hours count. Students needed about 40-60 hours together for a casual friendship and 80-100 for a friendship; meaningful talk, catching up and joking predicted getting closer, while more small talk predicted drifting apart (src:hall-2018).
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
friendship: from meeting in shared settings to a bond that acts on the world (reader's view)
- L1: Meeting: shared settings and nearness, the access to others that friendship needs
- L2: Recognising and attending: the other known as an individual, social stimuli rewarding
- L3: Choosing: by similarity (homophily), nearness and choice; strangers test the waters
- L4: Building the bond: hours together, talk, many small favours
- L5: Holding the bond: emotional bookkeeping, loose reciprocity, trust, stress buffered by the partner
- L6: Meaning: tacit expectations, the culture's category of friend, the tie as perceived
- L7: The pair acting on the world: help, alliances, support, influence, and the health and survival that follow
the boardroom (shared floors)
- L4: The shared history: bond strength built from repeated exchanges of help, hours and talk, as association builds in learning
- L5: Holding the bond: emotional bookkeeping, loose reciprocity, oxytocin and endorphins, stress buffered by the partner, trust
- L6: Meaning: the friendship as tacit expectations (positive regard, disclosure, aid, similarity, enjoyment, agency), as a culture's category and as a perceived tie
- L7: The pair acting: help, coalitions and alliances, support, influence on each other, ending or switching
way in: Emotion (inner): affection, reward, the pain of loss, jealousy
- L1: Social stimuli felt as intrinsically rewarding
- L2: Exclusion and loss felt akin to physical pain
- L3: Jealousy attuned to a rival who might replace one
way in: Hearing: talk, calls and self-disclosure
- L1: Everyday talk between acquaintances
- L2: Voices and calls recognised and remembered (dolphin whistles for up to 20 years)
- L3: Kinds of talk weighed: striving talk (meaningful conversation, catching up, joking, affection) builds closeness, small talk lowers it; problem talk can turn to co-rumination
way in: Help received: food, aid, support in fights
- L1: Food regurgitated to an unsuccessful roost-mate; aid and support given
- L2: Help remembered: monkeys respond differently to calls of those who groomed them
- L3: Imbalances tolerated rather than counted as cheating
way in: Interoception: one's own stress state
- L1: A stressor raising cortisol; the rise smaller with a friend present
- L2: Baseline cortisol lower in tightly knit networks
- L3: The stress response as part of the motivation that drives social contact
way in: Likeness: cues of shared traits, tastes and responses
- L1: Others who differ or match on age, gender, class and role
- L2: Responses to the world read as alike: friends' brains respond alike to films
- L3: Selection for similarity at the outset of a new setting
way in: Sight and presence: time spent near the other in shared settings
- L1: Shared settings and nearness: classes, dormitories, roosts, the same vicinity
- L2: The other recognised as an individual, by face in humans and macaques
- L3: Hours together tallied: more hours, closer friendship
way in: The society around: how freely relationships can be chosen
- L1: Relational mobility: a society's freedom to choose and drop relationships
- L2: Perceived as the opportunity and choice people see in their society
- L3: Who counts as a friend set by culture
way in: Third parties: friends of friends, others' relationships
- L1: Relationships between other pairs (friends of friends, friends of foes)
- L2: Third-party relations read: monkeys surprised when a known rank order is reversed
- L3: Friends who are not friends with each other weigh on girls' suicidal thoughts
way in: Thought (inner): beliefs about being liked and about the tie
- L1: Where thoughts about being liked come from: left open (contested origin); people underestimate being liked
- L2: Metaperception: estimates of what the other thinks of one; people underestimate being liked (the liking gap), for months
- L3: Ties assumed reciprocal; people poor at perceiving reciprocity
way in: Touch: grooming, huddling, contact
- L1: Grooming, huddling and contact between partners
- L2: Grooming received lowers heart rate
- L3: Oxytocin release after grooming only with a bonded partner
way out: Alliance: acting together against others (movement)
- L1: Estrus females herded away from their groups
- L5: Early associations become adult alliance partners
- L7: Allies recruited for conflicts
way out: Help: food, aid, support in fights (movement)
- L1: A blood meal regurgitated to a hungry roost-mate
- L5: Reciprocity loose, not immediate
- L7: Help given to a bonded partner
way out: Influence: shaping the friend's inner states and acts
- L1: A friend's learning rate changed
- L6: Influence passes only along ties that run the right way
- L7: Friends shape inner states, values and the acts that follow
way out: Leaving: ending, drifting or switching partner
- L1: Guilt, anger and depression in the one left
- L5: Switching to a new partner as a sanction
- L7: A friendship ends
way out: Talk: companionship and emotional support (speech)
- L1: Self-disclosure and social support given
- L7: Companionship, affection and emotional support
way out: Touch: grooming and huddling (movement)
- L1: The groomed partner's heart rate lowered
- L7: Grooming and huddling as affiliation
Sources (31) published science
Wikipedia: Friendship (read in full) (behind its links)
Zhou W.-X., Sornette D., Hill R.A., Dunbar R.I.M. (2005) Discrete hierarchical organization of social group sizes. Proc R Soc B 272:439-444. doi:10.1098/rspb.2004.2970 (behind its links)
Dunbar's number (Wikipedia, rev 1374122898, 2026-09-10) (behind its links)
Parkinson C, Kleinbaum AM, Wheatley T. 2018. Similar neural responses predict friendship. Nature Communications 9:332. doi:10.1038/s41467-017-02722-7 (PMC5790806, CC BY) (behind its links)
Carter GG (2024) Modeling the evolution and formation of animal friendship. PNAS 121(15):e2403318121. doi:10.1073/pnas.2403318121 (commentary, full text read) (behind its links)
Wikipedia: Homophily (behind its links)