Abstract
In the paradigm of Lewis-Skyrms signaling games, the emergence of linguistic conventions is a matter of equilibrium selection. What happens when an equilibrium has ``natural salience'' -- stands out as uniquely attractive to the players? We present two models. We find that the dynamics of natural salience can encourage the learning of more successful signaling conventions in some contexts, and discourage it in others. This reveals ways in which the supposed worst-case scenario -- a lack of natural salience -- might be better than some cases in which natural salience is present.
