Scalp electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) have highly complementary domains, and their combination has been actively sought within neuroscience research. The important gains in fMRI sensitivity achieved with higher field strengths open exciting perspectives for combined EEG-fMRI; however, simultaneous acquisitions are subject to highly undesirable interactions between the two modalities, which can strongly compromise data quality and subject safety, and most of these interactions are increased at higher fields. The work described in this thesis was centered on the development of simultaneous EEG-fMRI in humans at 7T, covering aspects of subject safety, signal quality assessment, and quality improvement. Additionally, given the potential value of high-field EEG-fMRI to study the neuronal correlates of so-called negative BOLD responses, an initial fMRI study was dedicated to these phenomena. The initial fMRI study aimed to characterize positive (PBR) and negative BOLD responses (NBR) to visual checkerboard stimulation of varying contrast and duration, focusing on NBRs occurring in visual and in auditory cortical regions. Results showed that visual PBRs and both visual and auditory NBRs significantly depend on stimulus contrast and duration, suggesting a dynamic system of visual-auditory interactions, sensitive to stimulus contrast and duration. The neuronal correlates of these interactions could not be addressed in higher detail with fMRI alone, yet could potentially be clarified in future work with combined EEG-fMRI. Moving on to simultaneous EEG-fMRI implementation, the first stage comprised an assessment of potential safety concerns at 7T. The safety tests comprised numerical simulations of RF power distribution and real temperature measurements on a phantom during acquisition. Overall, no significant safety concerns were found for the setup tested. A characterization of artifacts induced on MRI data due to the presence of EEG components was then performed. With the introduction of the EEG system, functional and anatomical images exhibited general losses in spatial SNR, with a smaller loss in temporal SNR in fMRI data. B0 and B1 field mapping pointed towards RF pulse disruption as the major degradation mechanism affecting MRI data. The main part of this work focused on EEG artifacts induced by MRI. The first step focused on optimizing signal transmission between the EEG cap and amplifiers, to minimize artifact contamination at this important stage. Along this line, adequate cable shortening and bundling effectively reduced environment noise in EEG recordings. Simultaneous acquisitions were then performed on humans using the optimized setup. On average, EEG data exhibited clear alpha modulation and average visual evoked potentials (VEP), with concomitant BOLD signal changes. In the second step, a novel approach for head motion artifact detection was developed, based on a simple modification of the EEG cap, and simultaneous acquisitions were performed in volunteers undergoing visual checkerboard stimulation. After gradient artifact correction, EEG signal variance was found to be largely dominated by pulse artifacts, but contributions from spontaneous motion were still comparable to those of neuronal activity. Using a combination of pulse artifact correction, motion artifact correction and ICA denoising, strong improvements in data quality could be obtained, especially at a single-trial level.