This work investigates the validity of the stick model used in diffusion-weighted MRI for modelling cellular projections in brain tissue. We hypothesize that the model will fail to describe the signals from grey matter due to an abundance of cell bodies. Using high b-value (≥3 ms/µm2) data from rat and human brain, we show that the assumption fails for grey matter. Using diffusion simulation in realistic digital models of neurons/glia, we demonstrate the breakdown of the assumption can be explained by the presence of cell bodies. Our findings suggest that high b-value data may be used to probe cell bodies.