Friday, 4 September 2026
H HiPerformance CultureThe Evidence
Social 5 min read
Trend Breakdown
The Evidence

Is Dunbar's number real?

Robin Dunbar's claim that human brains cap meaningful relationships at roughly 150 people has become shorthand for friendship advice and Slack channel sizing alike. The layered structure behind it is genuinely well observed, but the number itself sits inside a confidence interval wide enough to make it almost meaningless.

Published · 5 sources
Skip to the verdict ↓
Trend Science
Breakdown
Evidence-graded series

02What's being claimed

The claim is that the human brain is cognitively wired to sustain no more than about 150 stable relationships, a hard ceiling derived from the same neocortex scaling that predicts group size across primates. Push past it, the argument goes, and social cohesion quietly breaks down, which is why organisations from armies to tech companies have used it to cap team and community sizes.

Robin Dunbar arrived at the number by regressing group size against relative neocortex volume across 36 primate genera, then extrapolating the line to the human ratio, which predicted a mean group size of roughly 150 1. The mechanism sounded elegant: bigger brains, bigger groups, a single evolutionary dial anyone could point to. Malcolm Gladwell's The Tipping Point turned that regression line into a hard rule of thumb, stripping out Dunbar's original caveats along the way, and the figure has circulated ever since in management books and community-design guides.

Independent data sources kept landing in the same neighbourhood. Personal social networks really do show a nested, layered structure of roughly 5, 15, 50 and 150 people, distinguished by contact frequency and closeness 2, and an analysis of 1.7 million Twitter users found people sustain a maximum of around 100 to 200 stable, reciprocal relationships regardless of follower count 3. That convergence across brains, phone records and social media is what gives the claim its scientific shine.

Origin
Dunbar's 1992 study
Robin Dunbar regressed primate group size against neocortex volume, extrapolating a predicted human ceiling of about 150.
→
Vector
The Tipping Point
Malcolm Gladwell's 2000 bestseller popularised 150 as a hard cognitive and organisational limit, dropping the original caveats.
→
Spike
Remote-work community sizing
Post-2020 essays on 'organisational Dunbar numbers' revived the figure as a design heuristic for Slack and Discord communities.
"Your brain physically cannot handle more than 150 real relationships. It's called Dunbar's number, it's hardwired by your neocortex, and it's why your close-friends list should never grow past 150 no matter how big your following gets."
— Representative of the claim as it circulates online

03The evidence verdict

H
HiPerformance Culture The Evidence · Trend Breakdown
Verdict

The layered relationship structure is real; 150 as an exact cognitive limit is far shakier than the meme claims.

Moderate confidence evidence position 52/100 · 5 sources cited · 1 primate regression study, 1 narrative review, 2 large-scale observational studies, 1 statistical reanalysis · 1992-2021 · how we grade ↓
✓

What holds up

✓
Regressing group size against relative neocortex volume across 36 primate genera and extrapolating to the human ratio predicts a mean group size of about 150 1.
Gold
✓
Personal social networks show a consistent nested structure of roughly 5, 15, 50 and 150 people, set by contact frequency and closeness 2 4.
Silver
✓
An analysis of 1.7 million Twitter users found people sustain roughly 100 to 200 stable, reciprocal relationships regardless of follower count, matching Dunbar-scale predictions 3.
Silver
✕

What doesn't

✕
A 2021 reanalysis of Dunbar's original primate dataset using modern phylogenetic methods produced confidence intervals of 4 to 520 people, too wide to support deriving a precise cognitive limit 5.
Gold
✕
The popularised 5-15-50-150 scheme is a rounded simplification; phone-call records show actual layer boundaries (roughly 4-6, 12-15, 40-45) vary between individuals, not one fixed ratio 4.
Silver
✕
The Twitter study attributes its ceiling to finite time and attention, not an immutable, brain-hardwired cap, a distinction the popular '150 is your biological limit' framing elides 3.
Bronze

05So what do you actually do

If you use 150, treat it as a rough heuristic, not a biological constant.

The layered structure underneath it is the part worth keeping.

01Treat 150 as a loose order of magnitude when sizing a team, community or group chat, one input among several rather than a hard ceiling to defend.
02Prioritise the innermost 5 and 15 people in your life; those layers absorb almost all the maintenance time you actually have.
03Expect your own layer sizes to vary from the textbook 5-15-50-150 figures; individual variation is normal, not a sign you're doing relationships wrong.
04Judge a group by whether people in it actually still talk to each other, since contact frequency is what the layered structure is built from, rather than by its headcount.

04The studies

Scored on Design quality Measurement precision Causal clarity Replication value
GoldThe strongest evidence available for the specific line it supports: typically systematic reviews, meta-analyses, or randomised trials, or definitive mechanistic work where the line is a mechanistic claim.
SilverSolid but not decisive support: controlled or observational studies, or direct evidence that is small or not yet replicated.
BronzePreliminary or indirect signals; suggestive, never sufficient on their own.
Tiers are assigned by the evidence desk against the four axes above, judged relative to the claim each source supports; every source discloses its design and a resolvable DOI so the grade can be checked. A verdict never claims more confidence than its evidence base supports, and safety-critical findings are flagged at full contrast regardless of tier.
Gold
150 predicted human group size from primate neocortex regression
Comparative regression analysis · 36 primate genera
Dunbar, R. Journal of Human Evolution · 1992
Dunbar regressed group size against relative neocortex volume across 36 living primate genera, establishing a consistent scaling relationship between brain structure and typical social-group size. Extrapolating that regression line to the human neocortex ratio produced a predicted mean group size of approximately 150, the figure that later became shorthand for a hard cognitive ceiling on human relationships.
doi:10.1016/0047-2484(92)90081-j Verify ↗
Contested — a later phylogenetic reanalysis found the regression's confidence interval too wide to support deriving a single number
Silver
4 nested relationship layers identified
Narrative review · integrates psychological and evolutionary evidence
Sutcliffe et al. British Journal of Psychology · 2011
Sutcliffe and colleagues synthesised psychological and evolutionary evidence for a nested hierarchy of relationship layers, roughly 5, 15, 50 and 150 people, each defined by a different level of emotional closeness and contact frequency rather than by a single hard ceiling. The review argued the layered structure, not the number 150 alone, is the more robust finding.
doi:10.1111/j.2044-8295.2011.02061.x Verify ↗
Silver
100-200 stable relationships sustained per Twitter user
Large-scale observational · n=1.7M Twitter users, 6-month conversation data
Gonçalves et al. PLoS ONE · 2011
Analysing six months of conversational data from 1.7 million Twitter users, Gonçalves and colleagues found people maintained a maximum of roughly 100 to 200 stable, reciprocal relationships regardless of how many people they followed or were followed by. The ceiling broadly matched Dunbar-scale predictions, though the authors attributed it to finite time and attention rather than a fixed biological cap.
doi:10.1371/journal.pone.0022656 Verify ↗
Silver
~3x scaling ratio between adjacent relationship layers
Observational · mobile phone call-detail record analysis
Mac Carron et al. Social Networks · 2016
Mac Carron and colleagues analysed mobile-phone call-detail records to test the popularised 5-15-50-150 layering directly against real contact data. The layered personal-network structure held up, but the empirical layer sizes, roughly 4 to 6, 12 to 15 and 40 to 45 people, diverged from the popularised figures, with a scaling ratio of about three between adjacent layers rather than the assumed multiples.
doi:10.1016/j.socnet.2016.06.003 Verify ↗
Gold
4-520 95% confidence interval for predicted group size under reanalysis
Statistical reanalysis · phylogenetic generalised least-squares and probability-based modelling
Lindenfors et al. Biology Letters · 2021
Lindenfors and colleagues reanalysed Dunbar's original primate dataset using modern phylogenetic generalised least-squares methods, which account for shared evolutionary history between species far more rigorously than the original 1992 regression. The resulting confidence intervals were so wide, spanning 4 to 520 people, that the authors concluded a specific cognitive limit on human group size cannot be reliably derived from brain-size data alone.
doi:10.1098/rsbl.2021.0158 Verify ↗

06The verdict triad

Claim

Brains cap relationships at 150

The pitch: your neocortex hard-limits you to roughly 150 stable relationships, a number derived straight from primate brain scaling and echoed in phone records and Twitter data. That convergence across very different data sources is genuinely real, which is exactly why the claim travels so well.

Consequence

The exact number is shaky

The statistical derivation behind 150 is far weaker than its confident retelling suggests. A 2021 phylogenetic reanalysis of Dunbar's own primate dataset produced confidence intervals spanning 4 to 520 people, and real-world layer sizes measured from call records diverge from the popularised 5-15-50-150 figures.

Lever

Use the layers, not the number

Treat 150 as a rough order of magnitude rather than a precise cutoff, and focus instead on the better-supported layered structure: a handful of intimate ties, a wider circle of close friends, and a much larger outer band, each needing a different amount of maintenance time.

08What to do next

What to do next

Curious how your own relationship layers stack up?

Take the Social Connection Assessment to see how your inner circle, close friends and wider network compare with the evidence-backed layered structure.

Take the assessment

09Share & references

Update log
First published, based on 5 peer-reviewed sources spanning 1992-2021.
Machine layer added: structured claim data with an authored veracity verdict, accessible meter with stated evidence position, grading disclosure, and related links and calls to action wired to live destinations.
Next reviewDue by , or earlier if a new controlled trial or systematic review lands.
Related
Topic hub
Networks
Every evidence-graded piece on this shelf: guides, breakdowns, and instruments.
Browse the shelf →
Assessment
Social Intelligence Test
Find your limiting factor with an evidence-based self-assessment.
Start assessment →
Trend Breakdown
Networking Events
We grade the evidence behind the trend's biggest claims.
Read the breakdown →
Bibliography · every source, resolvable
01Dunbar, R. (1992). Neocortex size as a constraint on group size in primates. Journal of Human Evolution, 22(6), 469-493. doi:10.1016/0047-2484(92)90081-j Verify ↗Gold
02Sutcliffe, A., Dunbar, R., Binder, J. & Arrow, H. (2011). Relationships and the social brain: Integrating psychological and evolutionary perspectives. British Journal of Psychology, 103(2), 149-168. doi:10.1111/j.2044-8295.2011.02061.x Verify ↗Silver
03Gonçalves, B., Perra, N. & Vespignani, A. (2011). Modeling Users' Activity on Twitter Networks: Validation of Dunbar's Number. PLoS ONE, 6(8), e22656. doi:10.1371/journal.pone.0022656 Verify ↗Silver
04Mac Carron, P., Kaski, K. & Dunbar, R. (2016). Calling Dunbar's numbers. Social Networks, 47, 151-155. doi:10.1016/j.socnet.2016.06.003 Verify ↗Silver
05Lindenfors, P., Wartel, A. & Lind, J. (2021). ‘Dunbar's number’ deconstructed. Biology Letters, 17(5). doi:10.1098/rsbl.2021.0158 Verify ↗Gold
Build better performance.
Get the 90-day protocol.
Start the protocol