Directional Statistics: Computation and Interpretation of Histograms of Directional Data

Question

Suppose $v$ is a unit random vector in $R^n$ and consider a set $V = \{v^1,v^2,\dots\}$ of samples of $v$ gained either through experiments or simulations. How do we compute and interpret the histograms of $V$ given that $v$ is topologically in $S^{n-1}$? If we compute histograms of each component and display them on $R$ (call this a "naive histogram"), do we loose any geometry? If such histograms appear Gaussian when naively computed, do they correspond to wrapped normal distributions? If not how should a naive histogram be interpreted or a true histogram computed?