This paper is included in the Special Publication entitled 'What drives metamorphism and metamorphic reactions?' edited by P.J. Treloar and P.J. O'Brien. A long-standing problem is the extent of selective removal of chemical components and the amount of volume loss during the formation of slaty and crenulation cleavages during prograde metamorphism of sediments. Heterogeneous deformation leads to the formation of strongly deformed phyllosilicate-rich (P) domains and less deformed quartz-rich (Q) domains. The rocks generally show evidence of solution of quartz in P domains and addition of quartz in Q domains. The problem is whether the subtractions and additions balance or whether silica is lost from the system on the metre or broader scales. Structural studies tend to favour volume loss, whereas chemical studies tend to favour no volume loss. This review indicates that it is not yet possible to determine which of these views is correct. Deformation heterogeneity and fluid activity on a range of scales may play an important part in the development of slaty cleavage, suggesting the need for a detailed coordinated study, in which structural geologists and geochemists study identical material from the same sedimentary beds, in different parts of folds. Local compositional changes may develop adjacent to quartz veins, and some aluminous ('pseudopelitic') mineral assemblages may develop in igneous rocks by base-cation leaching before or during metamorphism. However, broad chemical changes with metamorphic grade are difficult to determine, owing to protolith variation on all scales, although much more detailed chemical work with good geological control remains to be done. Marked local protolith variability may also hamper attempts to determine the chemical effects of cleavage formation. As the density of metamorphic rocks increases generally with increasing grade, inferred conservation of mass on a regional scale implies loss of volume, which conceivably could be accommodated by collapse during deformation. Inferred volume loss implies a mass loss appropriate to the density increase. Inferred volume conservation implies regional mass gain, which is intuitively less probable. Marked chemical and volume changes occur when partial melts are removed at highest metamorphic grades.