The model of the casein micelle as a roughly spherical, fairly swollen particle of about 100 nm diameter with a hairy outer layer, is now generally accepted. Several workers have further assumed that the core of the micelle is built of sub-micelles, i.e., roughly spherical units of about 14 nm diameter, which are fairly tightly aggregated. It was originally assumed that the sub-micelles are held together in the micelle by bridges consisting of calcium phosphate. However, several studies, notably by Holt, have made this concept untenable. This also has made the existence of sub-micelles doubtful. Some years ago, however, van Dijk proposed that in the process of milk synthesis, casein molecules associate into small mixed clusters and that subsequently calcium phosphate is deposited into these particles, which would make them more dense (containing less solvent) and of a lower charge. Consequently, the particles would lose their colloidal stability and aggregate to form casein micelles. This concept agrees well with electron microscopic observations, notably by Buchheim, on milk formation in the lactating cell. It would also fit Brownian dynamics and molecular dynamics simulations of aggregation of small particles exhibiting weak colloidal attraction. Moreover, it readily fits several indications for a dynamic equilibrium between casein micelles and sub-micelles, for which examples of these will be given.