The recently measured dimensionless moment of inertia (MoI) factor for Callisto of 0.3549±0.0042 (Anderson et al., 2001, Icarus, 153, 157-161) poses a problem: its value cannot be explained by a model in which Callisto is completely differentiated into an ice shell above a rock shell and an iron core such as its neighboring satellite Ganymede nor can it be explained by a model of a homogeneous, undifferentiated ice-rock satellite. We show that Callisto may be incompletely differentiated into an outer ice-rock shell in which the volumetric rock concentration is close to the primordial one at the surface and decreases approximately linearly with depth, an ice mantle mostly depleted of rock, and an about 1800 km rock-ice core in which the rock concentration is close to the close-packing limit. The ice-rock shell thickness depends on uncertain rheology parameters and the heat flow and can be roughly 50 to 150 km thick. We show that if Callisto accreted from a mix of metal bearing rock and ice and if the average size of the rocks was of the order of meters to tens of meters, then Callisto may have experienced a gradual, but still incomplete unmixing of the two components. An ocean in Callisto at a depth of 100-200 km is difficult to obtain if the ice is pure H 2O and if the ice-rock lithosphere is 100 km or more thick; a water ocean is more plausible for ice contaminated by ammonia, methane or salts; or for pure H 2O at a depth of 400-600 km.