Institute for Mathematical Behavioral Sciences

Triadic configurations have been at the heart of theoretical and methodological advances in social networks for nearly half a century. Triads are implicated in many of social networks theoretical sacred cows -- structural balance, transitivity, random and biased nets, linkage between micro network properties and macro social structures, the strength of weak ties, structural holes, and network closure, to name a few. Triads and properties of triples also provide the basis for important social network methodologies the triad census, relational algebras, and structural effects in many exponential random graph models, for example. In this talk I replicate and extend results presented in my paper Comparing Social Networks: Size, Density, and Local Structure, which demonstrates that triad censuses for a wide range of social networks are largely accounted for by network density and dyadic distributions, properties more local than triads. That paper is available at http://www.socsci.uci.edu/~kfaust/papers/Comparing_Networks_121205.pdf The current talk presents results for a collection of 82 social networks, representing a number of different species (humans, baboons, macaques, bison, cattle, goats, sparrows, caribou, and more) and a variety of social relations (friendship, negative sentiments, choice of work partners, advice seeking, reported social interactions, victories in agonistic encounters, dominance, and affiliation). Using a two-pronged analysis strategy I show that more than 90% of the variation in triad censuses for these networks is accounted for by dyadic distributions, and that these triad censuses are virtually indistinguishable from what is expected given their dyad distributions. Implications of these results for social network methods and theories are discussed.