This paper explicitly separates word-of-mouth communication from observational learning in analyzing stock market participation decision.We achieve such separation by applying a theoretical framework and by incorporating market return information to historical participation pattern at the aggregate level.Our results show that there are significant effects of both types of social learning on participation decision,even after adding controls on supply-side effects,province-level and time-varying characteristics.We further demonstrate that our constructed measure of observational learning and word-of-mouth communication reflect the nature of these passive and active interpersonal channels accurately.These two forms of social learning are also shown to have differential impact in the Bull market and the Bear market.