Abstract In a previous paper, the theoretical possibility of reducing optical losses in fused silica by minor additions of an alkali metal oxide was considered. The problem concerns a competition between density fluctuation losses (which decrease on addition of the modifying component) and compositional fluctuation losses (which increase, due to a tendency towards subliquidus phase separation). In order to combat the latter disadvantage, it is desirable to probe the effects of an addition of a second minor constituent, such as Al 2O 3, which tends to reduce the forces driving the phase separation. However, an explicit theory of concentration losses in a three-component glass that is capable of making quantitative predictions has never yet been presented. As a preliminary to the primary goal of finding the optimum additives for reducing total losses in the above context, such a formal theory is set out in this paper, and its accuracy tested by comparison with existing scattering measurements on some sodium aluminosilicate glasses.