The resolution in conventional BOLD FMRI is considerably lower than can be achieved with other MRI methods, and is insufficient for many important applications. One major difficulty in robustly improving spatial resolution is the poor image quality in BOLD FMRI, which suffers from distortions, blurring, and signal dropout. This work considers the potential for increased resolution with a new FMRI method based on balanced SSFP. This method establishes a blood oxygenation sensitive steady-state (BOSS) signal, in which the frequency sensitivity of balanced SSFP is used to detect the frequency shift of deoxyhemoglobin. BOSS FMRI is highly SNR efficient and does not suffer from image distortions or signal dropout, making this method an excellent candidate for high-resolution FMRI. This study presents the first demonstration of high-resolution BOSS FMRI, using an efficient 3D stack-of-segmented EPI readout and combined acquisition at multiple center frequencies. BOSS FMRI is shown to enable high-resolution FMRI data (1 ؋ 1 ؋ 2 mm 3 ) in both visual and motor systems using standard hardware at 1.5 T. Currently, the major limitation of BOSS FMRI is its sensitivity to temporal and spatial field drift. BOLD FMRI has become an important tool in neuroimaging because it represents a good compromise between spatial resolution, temporal sampling, and breadth of coverage. Nevertheless, the achievable spatial resolution in conventional FMRI is poor relative to other MRI methods: Typical BOLD FMRI experiments acquire voxel volumes of 50 -100 mm 3 (e.g., 4 ϫ 4 ϫ 6 mm 3 voxels), while anatomical reference scans are regularly collected with voxel volumes of just 1-5 mm 3 . The coarse resolution of conventional FMRI is acceptable for many applications; however, these methods are inadequate to elucidate fine cortical architecture or study small gray matter nuclei. For example, small voxels are necessary to resolve the ocular dominance columns [1], or retinotopic [2] and somatotopic [3] maps. Additionally, the ability to distinguish deep brain nuclei would have a major impact on, for example, the study of memory [4] or pain [5].High-resolution BOLD FMRI is limited in part due to the coupling of functional contrast to sources of image artifacts and signal loss. Because BOLD detects signal dephasing caused by deoxyhemoglobin, functional contrast requires long echo times. Unfortunately, the use of long T E enhances signal dephasing from other sources of frequency heterogeneity, resulting in signal loss near susceptibility boundaries. Additional image artifacts are caused by the need for long readouts, which introduce distortion in frequency-shifted regions and, to a lesser extent, image blurring due to relaxation. Hence, robust functional contrast, which requires long T E and long readouts, is fundamentally at odds with image quality, which favors short T E and short readouts. This tradeoff of signal detectability for image fidelity is intrinsically tied to the use of a contrast mechanism that is based on signal dephasing, limiting the ac...