In functional magnetic resonance imaging, a rapid method such as echo-planar (EPI) or spiral is used to collect a dynamic series of images. These techniques are sensitive to changes in resonance frequency which can arise from respiration and are more significant at high magnetic fields. To decrease the noise from respiration-induced phase and frequency fluctuations, a simple correction of the "dynamic off-resonance in k-space" (DORK) was developed. The correction uses phase information from the center of k-space and a navigator echo and is illustrated with dynamic scans of single-shot and segmented EPI and, for the first time, spiral imaging of the human brain at 7 T. Image noise in the respiratory spectrum was measured with an edge operator. The DORK correction significantly reduced respirationinduced noise (image shift for EPI, blurring for spiral, ghosting for segmented acquisition). While spiral imaging was found to exhibit less noise than EPI before correction, the residual noise after the DORK correction was comparable In functional MRI (fMRI), a time-series of images is collected to monitor changes in the subject's physiology during a challenge, e.g., cognition, exercise, or contrast uptake. Invariably for such techniques, rapid acquisition such as echo-planar imaging (EPI) (1) or spiral imaging (2) is used in order to avoid loss of temporal information and to suppress artifacts from respiration and cardiac functions that are confounds to the challenge being monitored. In the case of blood oxygen level-dependant (BOLD) functional neuroimaging (3), which will be the primary focus of this work, an echo time (TE) comparable toT* 2 is used in order to maximize the sensitivity to small BOLD changes inT* 2 . Unfortunately, the relatively long TE also makes the acquisition sensitive to changes in resonance frequency and other factors, causing various artifacts.NMR phase shifts can arise from instrument instability, bulk motion, or from cardiac and respiratory functions (4 -7). Phase variations between segments cause ghosting and extra noise in the time-series, which degrades functional data and the resulting maps. Zero-order phase (nonevolving in time) corrections using navigator echoes (8 -10) or retrospective modeling of cardiac and respiratory modulations (11-13) have been employed to reduce these effects.A major effect from respiration is that the resonance frequency in the brain region can vary in time even when the head of the subject is immobile (14 -16). These firstorder phase variations (evolving linearly in time) result from respiration-driven movement of organs in the thoracic and abdominal cavities, as well as to changes in the oxygenation state of the respired gas. The field modulations tend to vary weakly in the inferior-superior direction in accordance with relative proximity to the chest, although the dominant effect is a global frequency shift in the whole brain (16,17). In single-shot EPI images, changes in resonance frequency result in positional voxel shifts predominantly in the phase encode ...