The mineral in bones and teeth is an impure form of hydroxylapatite (HAP), the principal impurity being 2–5 wt.% carbonate. This mineral dissolves during remodelling of bone and also in dental caries as a result of the action of acids produced by osteoclasts and by bacteria, respectively. In enamel, demineralization proceeds with preferential loss of carbonate relative to phosphate. Surprisingly, in the early stages, the demineralization is subsurface. In order to facilitate the understanding of physical chemical aspects of these processes, we have undertaken studies of demineralization in model systems. We give three examples here. The first two used scanning microradiography in which the specimen is stepped across a 10–30 μm diameter X-ray beam. Intensity measurements allow calculation of the mineral mass per unit area in the X-ray path through the specimen. In the first experiment, porous HAP sections were separated from a reservoir of acidic buffer by a column initially filled with water (the diffusion length) and scanned with the X-ray beam perpendicular to the axis of the diffusion length. The rate of total loss of mineral along each profile was calculated from the scans. The rate of demineralization fell as the diffusion length increased. We believe the explanation is that the rate- controlling step is the diffusion of dissolved HAP away from the solid to the buffer reservoir. In the second experiment, demineralizing solution and water were pumped alternately, for equal lengths of time, past blocks of porous HAP or enamel. The X-ray beam was perpendicular to the exposed surface. As the rate of switching between solutions decreased, the mean rate of demineralization also fell. We propose that this effect is due to retention of acid in the pores of the HAP during the time when water flows, allowing further demineralization to take place during this time. The third study used X-ray microtomography, a form of 3D microscopy, to study the loss of mineral in compacted carbonate apatite powders. The powders were packed in six 10 mm internal diameter acrylic cylinders to a depth of 4 mm (after pressing). One end was covered with a porous polyethylene disc and each tube placed in acidic buffer for 70 days. Periodic examination by microtomography showed the development of subsurface demineralization. Infrared spectroscopy of the dissected-out surface layers showed preferential loss of carbonate over phosphate by comparison with deeper layers. Rietveld analysis of X-ray powder diffraction data showed changes in the crystallographic structures of the apatites between the initial and dissected-out apatite.