Crop-to-crop gene flow in rice may affect the seed purity and quality of the recipients, in addition to other environmental consequences. Gene flow from cultivated rice to its wild relatives (crop-to-wild) may influence the genetic components and structures of the wild populations and even play a part in the extinction of the wild relatives. When transgenic rice is released into the environment for large-scale commercialization, the transgenes may escape from the transgenic rice into the nontransgenic rice varieties and its wild relatives, which might arouse serious concerns about biosafety.In order to estimate the frequencies of crop-to-crop and crop-to-wild gene flow in rice fields, different cultivated rice varieties including transgenic and nontransgenic rice were used and control field experiments were conducted to study the pollen-mediated gene flow in rice. Using the SSR (simple sequence repeat) molecular marker technology and hygromycin-resistant selection method for identifications of hybrids resulted from gene flow, this study examined totally more than 400 million rice seedlings.The main results and conclusions of this study were described as follows:(1) The frequencies of gene flow between different rice varieties were extremely low. The maximum gene flow frequencies detected in the experiments were all below 1%.(2) Outcrossing rate of pollen receiver was a limiting factor of rice gene flow. High outcrossing rate might lead to high gene flow frequency and the maximum frequency of gene flow was determined by outcrossing rate.(3) The pollen competitions at close spacing significantly affected the rice gene flow and gene flow frequency was in direct proportion to relative pollen density.(4) The gene flow frequencies dramatically reduced with the increase of the distance intervals between pollen donors and receivers. The results of this study showed that when the distance interval increased to 6 m and above, the gene flow frequencies dropped to less than 0.01%. Such results indicated that isolation distance could be an effective way to minimize crop-to-crop gene flow in rice.(5) In this study, there was no significant size effect on the gene flow frequencies,which might be because of the dramatic reduction of pollen density along distance. However, modeling of pollen-mediated gene flow indicated that size effect might become more and more observable according to the reduction of attenuation coefficient of pollen dispersal. When the size of pollen source increased, gene flow frequencies might increase as well, but the increase extent of the latter would reduce quickly and at last the gene flow frequencies might approach to limit points.(6) The study of the scale-effect showed that with the increase of the experimental scales the gene flow frequencies did not increase accordingly. Moreover, mathematical modeling showed that the gene flow frequency might decrease when experimental scale increased. Such results indicated that the little-scale experiments of rice gene flow could provide predictions for large-scale rice productions.(7) When there was only one pollen donor, gene flow frequency was in direct proportion to the pollen density.(8) The strong pollen competition from self-pollination might contribute to the very low outcrossing rates and gene flow frequencies in rice varieties.(9) The study of rice pollen flow showed that with the artificial disturbance of pollen dispersal there were pollen grains detected at 90 m from the pollen sources. Regression analysis indicated that the pollen grains could transfer more than 200 m far. Those results indicated that the rice pollen could dispersal very long distance.(10) Based on the experimental study, we designed a theoretical model for pollen-mediated gene flow in rice. The model could well represent the data of our study and so it could be used to simulate gene flow in rice fields.(11) The model simulations indicated that, when the outcrossing rate was no more than 1%, a 5 m buffer area with pollen receiver individuals between pollen donors and receivers was sufficient to minimize the frequencies of gene flow to <0.9%. However, for the crop-to-wild gene flow, the model showed that the frequencies of gene flow from cultivated rice to the adjacent wild rice populations could be considerably high and might transfer very long distance.This study can provide the scientific data for helping the establishments of the control measures of pollen-mediated gene flow in rice. Especially, to the crop-to-cropgene flow in rice, our model can be easily applied for the simulations and estimates. Our study can provide valuable references for the pollen-mediated gene flow in other self- and wind-pollination crops, such as wheat, for the pollen-mediated gene flow models of wind-pollination crops may follow the same principles.