Biophysical modeling is the mediator of evaluating the cellular structure of biological tissues using diffusion-weighted MRI. It is however the bottleneck of microstructural MRI. Beyond the complexity of diffusion, the current development is hindered by the fact that biophysical models heavily rely on diffusion-specific properties of diverse cellular compartments that are still unknown and must be measured in vivo. Obtaining such parameters by straightforward fitting is hindered by the degenerated landscape of the likelihood functions, in particular, the signal obtained for multiple diffusion directions and moderate diffusion weighting strength is not enough to estimate these parameters: different parameter constellations explain the signal equally well. The aim of this study is to measure the central parameter of white matter models, namely the intra-axonal water diffusivity in the normal human brain. Proper estimation of this parameter is complicated due to (i) the presence of both intra- and extra-axonal water compartments and (ii) the orientation dispersion of axons. Our measurement involves an efficient suppression of extra-axonal space and all cellular processes oriented outside a narrow cone around the principal fiber direction. This is achieved using a planar water mobility filter -- a strong diffusion weighting that suppresses signal from all molecules that are mobile in the plane transverse to the fiber bundle. Following the planar filter, the diffusivity in the remaining compartment is measured using linear and isotropic weighting. We find the specifically averaged intra-axonal diffusivity $D_0 = 2.25\pm 0.03{\,\rm \mu m^2/ms}$ for the timing of the applied gradients. Extrapolation to the infinite diffusion time gives $D_\infty \approx 2.0{\,\rm \mu m^2/ms}$. This result imposes a strong limitation on the parameter selection for biophysical modeling of diffusion-weighted MRI.