The determination of the physical density by Compton scatter densitometry is dependent upon the geometry used for the measurement. The diameter of the sample, the density range, and the scattering volume size and shape influence the relative electron density values obtained by such a system. This work presents the results of an experimental analysis of these parameters in which the importance of each is evaluated for samples of known density. A bias in the computed Compton density, due to multiple scattering and attenuation, is defined; and an expression relating the density bias to these geometrical parameters is presented. The importance of applying corrections to the computed physical density for samples of large diameter and/or high density (cortical bone tissue and large diameter cancellous bone tissue) is discussed.