Microscopic anisotropy was measured in vivo using a novel tensor-valued diffusion encoding approach. In gray matter, the microscopic anisotropy was generally low, but its variation corresponded well to known differences in myelination. We hypothesize that myelinated axons cause microscopic diffusion anisotropy but that the contribution from dendrites is negligible. This hypothesis is supported by comparisons with independent myelin assessments using T1W/T2W-ratios, T2-mapping, and myelin stains from histology. We also demonstrate that the “neurite density index” detected by NODDI is less sensitive to these changes, and why NODDI cannot map the neurite density accurately.