Grammars of action in human behavior and evolution

Distinctive human behaviors from tool-making to language are thought to rely on a uniquely evolved capacity for hierarchical action sequencing. Unfortunately, testing of this idea has been hampered by a lack of objective, generalizable methods for measuring the structural complexity of real-world behaviors. Here we present a data-driven approach for quantifying hierarchical structure by extracting action grammars from basic ethograms. We apply this method to the evolutionarily-relevant behavior of stone tool-making by comparing sequences from the experimental replication of ~2.5 Mya Oldowan vs. more recent ~0.5 Mya Achuelean tools. Results show that, while using the same 9alphabet9 of elementary actions, Acheulean sequences are structurally more complex. Beyond its specific evolutionary implications, this finding illustrates the broader applicability of our method to investigate the structure of naturalistic human behaviors and cognition. We demonstrate one application by using our complexity measures to re-analyze data from an fMRI study of tool-making action observation.