Cross-modal learning refers to learning that involves obtaining information from multiple modalities and then integrating and utilizing it. Multisensory integration is an important basis of cross-modal learning. Although cross-modal conditions are more like the real-life environment of human learning, most studies still use single-modal stimuli, and the results of cross-modal learning are still somewhat messy and unsystematic. In order to better summarize the characteristics of cross-modal learning and its related mechanisms, the current review first introduces multisensory integration effects and the factors that influence them, as well as the experimental and theoretical researches on the modality non-specificity of primary cortexes. Then, we sum up the researches on consciousness, representation type, and transfer effect in cross-modal learning, and research progress for neural mechanism of cross-modal learning by adopting the techniques such as neuronal recording, ERP and fMRI. Finally, we summarize the current research results of cross-modal learning, and examine prospects for the potential applications of these research findings and future research issues in this area.