The results of a continuing series of laboratory experiments, designed to model the fluid motions which accompany crystallization, are both described and related in a preliminary way to prototype flows in magma chambers. Previous experiments have demonstrated the importance of compositional inhomogeneity, produced by crystallization and melting in a thermal gradient and coupled with double-diffusive effects, in driving convective flows which result in thermal and compositional stratification in an originally homogeneous fluid. The present experiments examine effects produced in tanks cooled at the side, by the upward flow of a less dense boundary layer depleted in the crystallizing component as crystals grow on the side wall. These processes are examined in simple two and three component aqueous systems (H 2O-Na 2CO 3, H 2O-Na 2CO 3-K 2CO 3, H 2O-CuSO 4-Na 2SO 4) with one and two crystallizing phases. In each of these systems, an initially downward flow of a cooled boundary layer against the side wall is reversed as crystallization commences and depletes the boundary layer in the crystallizing component. Accumulation of this cooler but lighter depleted fluid at the top of the chamber produces thermal and compositional layering by a “filling box” mechanism, partly modified by interchange between the boundary layer and the convecting layers outside. When more than one component is present in the solution, the crystallization process produces a differentiated fluid column, i.e. one with compositional gradients which are different for each of the components. The compositional and thermal distributions within the fluid change with time, but finally appear to reach a steady state. These distributions are the integrated result of compositional changes produced by crystallization from a thin boundary layer, a small proportion of the bulk fluid which evolves in composition and temperature independently of the bulk fluid, in a manner controlled by the dynamics of the system. The paths of fluid evolution, and the resulting sequence and abundance of crystalline products, are very different from those predicted on the basis of simple thermally-driven crystallization in the static system. Disequilibrium effects, probably involving different kinetics of crystallization of different crystals, in part determine crystallization in the experimental system; these may or may not be important in real magmatic systems. There is still much to learn before even the behaviour in these simple experimental systems is understood fully, and the application of the results to the interpretation of the behaviour in real magma chambers in even further away. However, it is apparent that dynamic effects provide an additional degree of freedom in crystallization processes, and allow evolution of small increments of solution to compositions much further along crystallization paths than that allowed in static equilibrium systems. Petrologists will have to free their thinking from classical constraints of static systems in order to interpret many magmatic phenomena.