Rumor spreading is a typical social phenomenon. Therefore, it is essential to learn the evolution mechanism of rumor spreading for social networks. In comparison with the classical SIR rumor spreading model, we introduce a new stochastic model for the spread of rumors, called Spreader-Ignorant-Recaller-Stifler (SICR) model, by adding a new recall group of people who stop spreading the rumor temporarily and could spread it again when incentives occur. Then mean-field equations are derived to describe the dynamics of the model. Numerical solutions of the new model are used to analyze the impact factors of rumor spreading through changing the parameters.