Abstract: Objective To evaluate the effect and mechanism of different frequencies of repetitive transcranial magnetic stimulation （rTMS） on speech function recovery in patients with moderate to severe motor aphasia after stroke. Methods Thirty patients with moderate to severe motor aphasia after stroke were recruited in this study and divided into the control group, 1 Hz rTMS （low-frequency rTMS, LF-rTMS） group and 10 Hz rTMS （high-frequency rTMS, HF-rTMS） group, 10 cases in each group. All patients received regular language training. Patients in the LF-rTMS and HF-rTMS groups received rTMS treatment at the corresponding frequency of the right inferior frontal gyrus triangle, followed by regular language training. The Chinese version of Western Aphasia Battery （WAB） and task state fMRI were performed before and after corresponding treatment to compare the differences in language level and the changes in brain region activation and activation voxel indices （AVI）. Results All patients showed significant improvement in auditory comprehension, repetition, aphasia quotients （AQ） after treatment （all P < 0.05）. Moreover, spontaneous speech and naming were significantly improved in both LF-rTMS and HF-rTMS groups （all P < 0.05）. Compared with the control group, repetition, naming and AQ were remarkably improved in the LF-rTMS group （all P < 0.05）, whereas spontaneous speech, auditory comprehension, naming and AQ were significantly improved in the HF-rTMS group （all P < 0.05）. Compared with the LF-rTMS group, the improvement in spontaneous speech and auditory comprehension in the HF-rTMS group was more significant （both P < 0.05）. The hemispheric dominance was shifted from the right to the left hemisphere after treatment in all three groups. fMRI showed either increased or decreased activation among both language-related regions of interest （ROI） and non-ROI after treatment, whereas the ROIs related to language perception and semantic comprehension and expression in the left hemisphere were the most activated. In the LF-rTMS group, the AVI was significantly increased in the left hemisphere （P < 0.05）. The left pars triangularis failed to show increased activation after suppressing the homologous region. In the HF-rTMS group, the AVI was significantly increasedin bilateral hemispheres （both P < 0.05）. The language-related ROI activation was increased in bilateral pars triangularis. Conclusions Both LF-rTMS and HF-rTMS can facilitate the speech function recovery in patients with moderate to severe motor aphasia after stroke, and the HF-rTMS seem to be more effective compared to LF-rTMS. LF-rTMS can optimize the language organization through greater activation of high-efficiency language function area in patients with post-stroke aphasia. HF-rTMS can enhance the compensatory effect of the right hemisphere and play a role in facilitating the reconstruction of language performance in the left hemisphere.