A model has been suggested by Haller and others for the development of interconnected submicrostructures in glasses which involves the nucleation, growth and coalescence of discrete second-phase particles. A theoretical objection has been raised to this model—viz., that interparticle interference effects limit diffusion controlled growth and prevent coalescence. The present paper discusses three mechanisms whereby coalescence may occur despite interference effects. These include heterogeneous nucleation between two particles; capillarity-induced volume diffusion; and diffusion driven by the variation with separation of the interfacial energy of two interfaces. Two theories of the latter variation are presented, a diffuse interface theory based on the variational method of Cahn and Hilliard, and a pair interaction, sharp-interface model. Both models predict a monotonic decrease in the interfacial energy as the separation decreases. Other aspects of coalescence as well as the alternative model of spinodal decomposition are briefly discussed. The results of various experiments attempting to distinguish between spinodal decomposition and nucleation-growth-coalescence are reviewed. It is concluded that the operative mechanism for the formation of interconnected submicrostructures in nearly all glass systems remains to be established.