[jello]

Also relevant is the follow-up paper published the following year, which extends Dunbar's argument:

Dunbar, R. I. M. 1993. Coevolution of neocortical size, group size and language in humans. Behavioral and Brain Sciences 16:681-735.

pdf: http://www.cogsci.ucsd.edu/~johnson/COGS184/3Dunbar93.pdf

That said, Dunbar's paper - and the entire concept of the "Dunbar number" - is the shoddiest piece of science I have ever seen. Dunbar arrives at the magical 150 by extrapolating from the neocortex ratio to mean group size correlation of nonhuman primates. However, the human neocortex ratio is 30% higher than that of any primate and the resulting "corresponding" mean group size of 147.8 is more than double the highest nonhuman mean group size of 65.5, which makes the validity of such a large extrapolation highly suspect. In an attempt to justify this, Dunbar gives a 95% confidence interval of 100.2 - 231.1 (quite a wide range), then sets out to find as many examples of real-world data as possible that fit in this predetermined interval. As is often the case when one knows exactly what to look for, it appears to be everywhere. My favorite moment is where he encounters Naroll (1956)'s observed maximum settlement size of 500 yet deems it "likely that the equivalent mean settlement size will not be too far from the value of 150 suggested by the above analyses" (Dunbar 1993 p687).

How are you supposed to go about falsifying an argument like that?

Edit: Better formatting.