Language Evolution and Computation Bibliography

In his book Grooming, Gossip and the Evolution of Language, biologist Robin Dunbar (1997) proposes a new way of looking at the evolution of language. According to this view, language evolved to provide a new social bonding mechanism: Gossiping. This allows humans to live in larger groups than other primates, which increasing predation risks forced our ancestors to do. We use a computational multi-agent model to test the internal workings of this hypothesis, with interesting results. Our work provides a fundamentally new kind of evidence for Dunbaras theory, by experimentally demonstrating that greater group sizes can stimulate the evolution of language as a tool for social cohesion.

Anthropologists and archaeologists increasingly use phylogenetic methods to test hypotheses involving cross-cultural traits, but the appropriateness of applying tree-based methods to analyze cultural traits is unclear. The authors developed a spatially explicit computer simulation model to investigate trait evolution in relation to phylogeny and geography and used the simulation to assess the sensitivity of two comparative methods (independent contrasts and partial Mantel tests) to different degrees of horizontal transmission. Simulation results show that (a) the method of independent contrasts is sensitive to even small amounts of horizontal transmission in cultural data sets, (b) Mantel tests fail to cleanly discriminate between datasets characterized by different levels of horizontal and vertical trait transmission, and (c) partial Mantel tests do not produce markedly improved statistical performance when testing for associations among traits (as compared to independent contrasts). The results highlight the need for empirical estimates of horizontal transmission and extinction rates in cross-cultural datasets.