The Compton scattering of X-rays within a volume of a sample depends on the density of that voxel. The potential was examined of imaging the density variation across a food material by measuring the Compton scatter pro"le across polystyrene spheres with internal voids. Particular attention was paid to the obscuring in#uence of multiple scatter. Polystyrene was used as a sample material as it has a density that is typical of a food material can be machined to convenient shapes. A Monte Carlo computer simulation was used to predict the contrast, signal-to-noise ratio and multiple scatter fraction of the Compton scatter signal for various combinations of beam energy, beam size, sample size and void size. The simulation study was found to be in close agreement with experimental observation. Scattering produced better contrasts than transmission. However, the signal-to-noise ratio from scattering was poor compared with transmission, because of increased statistical noise due to the smaller number of photon counts. The minimum time required to detect the void inclusions using a typical X-ray source was estimated and conclusions for food inspection applications were discussed. Voids could be detected in 1 s if their diameter was greater than th that of the sample. The simulation study indicated that the multiple scattered #ux was small compared to the primary scattered component and consequently did not signi"cantly in#uence the imaging of density discontinuities. This conclusion was substantiated by experimental measurement. Transmission and scatter imaging were compared for inhomogeneous samples.