Effective coordination is a key social ingredient and social structure may be approximated by networks of contacts. Using Stag Hunt games, which provide socially efficient and inefficient equilibria, we compare our simulation results using artificial players and evolutionary game theory with laboratory experimental work with human subjects on small-world type networks and with theoretical results. The conclusion is that the apparently encouraging results obtained in the few human experiments in which the local interaction structure seems to promote efficient equilibria, is neither supported by simulation results nor by theoretical ones.