The social brain meets neuroimaging Robin I.M. Dunbar Institute of Cognitive & Evolutionary Anthropology, University of Oxford, 64 Banbury Rd, Oxford OX2 6PN, UK Recent neuroimaging studies in humans have indicated that individual differences in social network size corre- late with amygdala volume and the volume of brain regions associated with theory of mind. A new article demonstrates that this is also true for monkeys. Taken together, these findings provide crucial support for the social brain hypothesis. The social brain hypothesis was first proposed (as the Machiavellian Intelligence Hypothesis) in the late 1980s to explain why primates have unusually large brains for body size compared to other vertebrates [1]. Primates live in unusually complex societies and, therefore, need bigger ‘computers’ to handle that complexity. In the two decades since then, considerable behavioural evidence has demon- strated that primate social complexity (indexed variously as social group size, the complexity of male mating strate- gies, the use of tactical deception and the use of coalitions) correlates across species with brain size (or more particu- larly with neocortex size) [2]. What this claim lacked, however, was evidence that differences in brain volume really matter socially. Some evidence that such a link might be involved derives from the finding that, in humans, the size of personal social networks correlates with mentalising competences: Stiller and Dunbar [3] demonstrated that performance on high order (multi-level) mentalising tasks correlated with the number of close friends that individuals had. This raised the possibility that there might be a three-way correlation between behaviour (network size), the ‘software’ (cognitive skills) and the ‘hardware’ (some aspect of brain volume). Four recently published neuroimaging studies, culminat- ing in a seminal paper by Sallet and colleagues published in Science in November 2011 [4], now provide direct evi- dence for this. The first of these was published in 2010 and demonstrat- ed a correlation between amygdala volume and social skills in adult humans. Bickart et al. [5] used indices that indi- rectly measured network size and complexity in a sample of 58 adults and found that both correlated with amygdala volume (but not with the hippocampus as a control region for contrast). They focussed on the amygdala because (i) it is credited with being involved in emotional coding of social signals and (ii) its basolateral nucleus has been shown to correlate with social group size in primates. In an explor- atory analysis with a more lenient threshold ( p<0.01 uncorrected), Bickart et al. also found significant correla- tions between network size and complexity and cortical thickness in some regions in the temporal and frontal lobes, as well as the anterior cingulate cortex (although these did not survive correction for multiple comparisons given the large number of analyses they undertook exploring the whole cortex). Although the amygdala is potentially of interest in this respect (and its basolateral nucleus does have a direct connection with the frontal cortex), the weak- ness with this study is simply one of scale: the amygdala is at best a volumetrically tiny unit and of itself cannot explain the cross-species correlations with total neocortex (or brain) volume. Something still seemed to be missing. The answer has now been provided by three new imag- ing studies. Kanai et al. [6] used voxel-based morphometry (VBM) in a large sample of adult humans to explore the relationships between several indices of sociability (the number of friends on Facebook, and indices of sociability, such as who came to your birthday party or of whom you could ask a favour) and the volume of the amygdala and several cortical regions (left middle temporal gyrus, right superior temporal sulcus, right entorhinal cortex) previ- ously identified as being associated with theory of mind competences. They reported significant correlations with all three brain regions and number of Facebook friends, but only correlations between temporal lobe regions and their sociality indices. These findings are in broad agreement with a second VBM study, which demonstrated a direct three-way corre- lation between social network size, mentalising skills (indexed by the ability to solve the Stiller and Dunbar [3] multilevel false belief tasks) and grey matter volume in the temporoparietal junction and the temporal pole [7]. However, there appeared, in particular, to be a shared neural substrate for both mentalising competences and network size in the medial orbitofrontal cortex and the ventromedial frontal gyrus. In fact, a gross stereological analysis of the same dataset had previously demonstrated that individual differences in mentalising competences (on a scale ranging from 2 nd to 5 th order intentionality) corre- lated with the volume of grey matter in the orbitofrontal cortex in the same subjects [8]. In effect, taken together, these studies confirm for the first time that the social brain hypothesis not only applies between species, but it also applies within species. This is important from an evolu- tionary view because it confirms that there is individual variation on which natural selection can operate (a pre- requisite for evolution to work). However, these studies all focussed on humans, whose social cognitive competences, as well as social group sizes, are obviously in a league of their own. Lurking behind these results was the question as to whether this result was peculiar to humans. Sallet et al. [4] have now provided the answer, using MRI with rhesus macaques to demon- strate that individual differences in the volumes of grey matter in the superior temporal sulcus and the rostral Spotlights Corresponding author: Dunbar, R.I.M. (robin.dunbar@anthro.ox.ac.uk). 101