Flint pebbles that have been worn down enough to be symmetrical may be spherical, or may approximate to prolate or oblate spheroids or to ellipsoids. When, however, the section is at all elongated, it is not as a rule accurately elliptical, but except at the axial points it lies entirely outside an ellipse adjusted to the same axes. Thus, if one of the axes is much smaller than the other, the pebble is much flatter than an ellipsoid. Considering the quasi-spheroidal pebbles, whether prolate or oblate, these are always flattened at the poles, and the very oblate ones become tabular or even concave at the poles. These statements hold for flint pebbles, but a large quartzite pebble is figured which is pretty accurately an oblate spheroid, the meridional sections being truly elliptical. Experiments are described with chalk pebbles, initially shaped as prolate or oblate spheroids, and the change of figure under abrasion is observed. This depends in some degree on what is the abrasive. Steel nuts, nails ('tin tacks’) and small shot were used. In general the axes tend to approach equality, but not rapidly enough for the spherical form to be attained before the pebble has disappeared. The form initially spheroidal becomes flattened at the poles just like the natural flint pebbles, and may become concave, as flints sometimes do. The circumstances determining the rate of abrasion at any point are considered, and it is shown that this abrasion cannot be merely a function of the local specific curvature. The figure at points other than the one under consideration comes into question, so that the problem in this form has no determinate answer. It is shown by simple mechanical considerations how the concave form arises.