Objective Conventional time-series parameters are unreliable descriptors of motor-evoked potentials (MEPs) in brain tumor patients. Frequency domain analysis is suggested to provide additional information about the status of the cortico-spinal motor system. Aim of the present study was to describe the time-frequency representation of MEPs and its relation to the motor performance. Methods This study enrolled 17 consecutive brain tumor patients with impaired dexterity. After brain mapping of the affected (AH) and non-affected (NAH) hemisphere, TMS was applied to the hotspots of the abductor pollicis brevis muscles (APB). Using a Morlet wavelet approach, event-related spectral perturbation (ERSP) and inter-trial coherence (ITC) of the MEPs were calculated and compared to the Grooved Pegboard Test (GPT). Additionally, inter- and intra-subject reliability was assessed by the intraclass correlation coefficient (ICC). Results MEPs were projecting to a frequency band between 30 and 400 Hz with a local maximum between 100 and 150 Hz. There was a significant ERSP and ITC reduction of the AH in comparison to the NAH. In contrast, no interhemispheric differences were depicted in the conventional time-series analysis. ERSP and ITC values correlated significantly with GPT results (r = 0.35 and r = 0.50). Time-frequency MEP description had good inter-and intra-subject reliability (ICC = 0.63). Conclusions Brain tumors affect corticospinal transmission resulting in a reduction of temporal and spectral MEP synchronization correlating with the dexterity performance. Significance Time-frequency representation of MEPs provide additional information beyond conventional time-domain features.