Cortical layer-dependent high-resolution fMRI can help understand the effects of afferent and efferent functional connectivity on brain activity at different cortical layers [1]. Because of known differences in input-output characteristics for anatomically defined cortical layers, we hypothesize that individual layers should show different resting-state signal fluctuations [2]. Here, we report on a new fMRI acquisition and analysis method that might be able to measure layer-dependent resting-state activity fluctuations. In addition, we show effective functional connectivity of layer-dependent resting-state signal fluctuations in the human sensory-motor system.Experiments in four volunteers were performed on a 7T Siemens scanner with a 32-channel NOVA Medical head coil. We used a multi-contrast VASO-BOLD sequence with slices positioned to be nearly perpendicular to the cortex to capture M1 and S1 of both hemispheres without aliasing artifacts in the A-P direction (Fig. 1A). Sequence parameters (similar to [3]), TI1/TI2/TR = 1.1/2.6/3.0 s, 7 slices with GRAPPA-2 and FLEET [4] without partial-Fourier imaging, rectangular imaging matrix with 140 phase encoding steps and 70 data points. Two functional scans were collected per subject: (1) 12 min long rest, (2) 12 min long unilateral finger-tapping (alternating 30 s rest and 30 s act). Physiological noise correction was performed with RETROICOR based on acquired respiration and cardiac traces. Segmentation and cortical layering algorithms (based on the equi-volume approach [5]) were implemented in C++ and applied directly on the EPI images from the functional scans (Fig. 1). In order to obtain smooth surfaces, EPI data were rescaled by expanding k-space by a factor of 4 with zero-filling (blue curved arrow in Fig. 1B-C). This allows us to model 10-15 layers across the cortical depth (4 mm in M1; color edges in Fig 1D). See also Fig. 2I for an example of smooth surfaces in rescaled EPI space.Functional VASO EPI time-series provide sufficient T₁-weighted contrast to apply surface/layering algorithms [5] directly on the functional data (Fig. 1D), obviating the need for coregistration and distortion-correction. tSNR of VASO (18) is about 48% of BOLD tSNR (38). However, CBV-weighted functional activity (purple in Fig. 1F/H) shows higher specificity to GM compared to BOLD (orange in Fig. 1E/G) for finger-tapping and specific ICA components (manually selected based on its position in M1). Placing seeds in individual layers of M1 shows distinct layer-specific correlation with numerous brain regions within the FOV (Fig. 1I). Layer-dependent correlation matrices of right-hand M1 with itself (Fig. 2A/E), as well as with ipsilateral S1 (B/F), contralateral M1 (C/G) and contralateral S1 (D/H), confirm that CBV-based fMRI reveals layer-dependent features, which are blurred and shifted towards the cortical surface in BOLD-fMRI.The higher localization specificity of CBV-fMRI compared to BOLD-fMRI is consistent with previous task-based VASO studies [1,3]. Specific input into superficial or deeper cortical layers of M1 is different between the individual brain areas. While input from primary somato-sensory areas correlates most strongly with upper layers in M1 (yellow arrows in Fig. 1I), the input from premotor regions is mostly correlated with deeper layers (blue arrows in Fig. 1I). The cross-correlation matrices (Fig. 2A-G) show layer-dependent features in M1, but not in S1. The presence of layer-dependent features in M1 but not in S1 suggests that the nominal in-plane resolution of 0.9 mm is sufficient to reveal layer-dependent resting-state activity in thick cortices (e.g. M1, 4 mm), but may not be sufficient for thinner cortices (e.g. S1, 1.8-2 mm). However, it might also be the case the individual layers in S1 are more synchronized, compared to M1. The fact that S1 projects its input into M1 mostly in upper layers in M1 (black ellipse in Fig. 1I and 2F) with a secondary peak in deeper layers (blue ellipse in Fig. 2F) is consistent with MEMRI tracer studies in rats [6].We were able to show layer-dependent resting-state signal features with CBV-based fMRI in humans. The higher specificity of CBV fMRI reveals layer-dependent signal variations more clearly than BOLD. We could identify that both M1 areas have the same source of resting-state signal fluctuations in their middle layer (green lines in Fig. 2), which is anticorrelated with the primary sensory areas (blue lines in Fig. 2).