'Dunbar's number' deconstructed.
Level 5 - mechanism / opinion, no new human data
Level 5 by design analogy; statistical re-analysis and comparative phylogenetic modeling using non-human primate datasets without direct human empirical data.
PubMed 33947220 · doi:10.1098/rsbl.2021.0158
What was done
The authors tested the statistical validity of 'Dunbar's number' (the theoretical ~150-person limit on human social groups derived from regressing primate relative neocortex size against group size). They applied Bayesian and phylogenetic generalized least-squares methods to complementary primate datasets to re-estimate predicted human group sizes.
What was found
Different analytical methods yielded widely differing average group size predictions: Bayesian methods estimated averages between 69 and 109, while generalized least-squares phylogenetic methods estimated averages between 16 and 42. Both approaches generated wide 95% confidence intervals (4–520 and 2–336, respectively), demonstrating that a single cognitive limit cannot be reliably identified.
Why it matters
This study demonstrates that the foundational regression analyses used to support Dunbar's number lack statistical precision, showing that primate brain scaling cannot establish a specific human social group size limit.
Limits
The abstract does not specify the exact number of primate species, sample sizes, or specific datasets analyzed. The findings address the validity of the cross-species anatomical regression method rather than measuring human social network capacity directly through empirical sociological or cognitive testing.
Cited by
- contradicts Dunbar's number is 150, representing the approximate limit of people with whom a human can maintain stable social relationships.