Models of regular cellular-solids representing femoral head 'medial group' bone were used to (1) compare thickness data for plate-like and beam-like structures at realistic surface areas and densities; (2) test the validity of a standard formula for trabecular thickness (Tb.Th); and (3) study how systematic changes in cancellous bone thicknesses, spacing, and face-connectivity affect relative density and surface area. Models of different face-connectivities, produced by plate removal from the unit cell, were fitted to bone density and surface area data. The medial group bone was anisotropic: the supero-inferior (SI) direction was the principal direction for bone plate alignment and the plane normal to this had the largest number of bone/void intersections per unit line length (P(I)). A comparison of boundary perimeter per unit area data, in planes normal to SI, with surface area data placed the medial group bone between prismatic structures in which walls are parallel to one principal direction and isotropic structures. Selective removal of plates from a closed-cell model produced a similar result. For the same relative density and surface-area, plate-like models had significantly thinner cross-sections than beam-like models. The formula for Tb.Th produced overestimates of model plate thickness by up to 20% at realistic femoral cancellous densities. Trends in data on surface area to volume ratio and density observed on sampled medial group bone could be simulated by plate thickness changes on models of intermediate face-connectivity (approximately 1.5) or by plate removal from models with relatively thick and short (low aspect-ratio) plates. The latter mechanism is unrealistic for it resulted in beam-like structures at low 'medial group' densities, an architecture unlike the predominantly plate-like bone in the sample.