Previous reports of reproducibility of seed-based resting state functional connectivity (rs-fMRI) have shown a wide range of reproducibility [1,2,3], impacted by scan length, physiologic noise, and choice of network measures. Though signal fluctuations at 7T are also impacted by sources other than neuronal activity, the percent of signal fluctuation attributable to neuronal activity is also increased [4]. Here we evaluate scan-rescan reliability of seed-based rs-fMRI 7 tesla.Five healthy controls and 3 patients with multiple sclerosis (MS) were scanned, repositioned and re-scanned on a Siemens 7T Magnetom MRI scanner with SC72 gradient (Siemens Medical Solutions, Erlangen) using a 32-channel head coil (Nova Medical). Rs-fMRI data were acquired using slice-accelerated EPI sequence (WIP 770, Siemens). The scan parameters were as follows: 132 repetitions of 81 slices acquired with TE/TR=21ms/2800ms, 1.5mm slice thickness, matrix 160x160, FOV 210mm x 210mm, multi-band factor = 3, grappa factor = 2, and receive bandwidth = 1562 Hz/pixel. Subjects were instructed to keep their eyes closed during scans. Scans were corrected for motion and physiologic noise, detrended, and lowpass filtered [5,6]. Using a previously published method [7], 9-voxel in-plane seeds were placed to evaluate the strength of connectivity in the transcallosal motor pathway (bilateral primary motor cortex) and the posterior portion of the cingulum bundle (posterior cingulate cortex and hippocampus). Correlation measures were normalized to z-scores using whole-brain correlation maps [8].Scan 1 and Scan 2 rs-fMRI scores in the transcallosal motor pathway were highly correlated (Figure 1; r = 0.937, p = 6 x 10-4; ICC = 0.904), as were test-retest measures of the right and left posterior cingulum bundle (r = 0.8567 p = 0.007; ICC = 0.729). Figure 2 shows individual whole-brain rs-fMRI maps in one subject at Scan 1 and Scan 2. Average percent difference from Scan 1 to Scan 2 was -8.9% in the motor cortex (range: -28.2 to 1.4%) and 4.4% in the PCC (range: -18.6 to 24.5%).The seed-based reliability presented here represents a “best-case” scenario, utilizing subjects who are acclimated to the scanning environment and produce little motion compared to a patient population. Though we find a high correlation between Time 1 and Time 2, within-subject variation was high in some subjects. Previous reports of periodicity in the rs-fMRI signal may contribute to this variation [9].Scan-rescan reliability of rs-fMRI at 7T was high in this sample, though some subjects showed high levels of variability across scans.