This paper is concerned with the interpretation of granitoid microstructures, especially microstructural evidence that granitoids crystallize from liquids, and that their grain shapes do not change appreciably during cooling, in the absence of deformation and metamorphism. The paper is prompted by persistent assertions that (1) granitoids contain abundant solid material (restite) from the source, (2) K-feldspar megacrysts grow very late, in largely or completely crystallized rock, and (3) granitoids undergo essentially metamorphic changes as they cool. I present microstructural evidence to argue against these three hypotheses. For an explanation of my use of the term 'microstructure' (= 'texture' for many petrologists), see Vernon [2004, p. 7].

Though some geochemists appear to regard chemical data as being more reliable than structural evidence, considerations of both fundamental properties of solids, structure and chemical composition, are important for making genetic interpretations. A problem with a whole-rock chemical analysis, used alone, is that it averages grain-scale chemical variations, just as a whole-rock isotopic composition averages the isotopic compositions of all the minerals in a rock, which can have very different magmatic histories [e.g., Davidson et al., 2007]. For example, crystals can grow in other parts of a magma chamber or another magma body, after which they can be mixed with crystals formed in situ, thereby becoming ‘antecrysts’. Fortunately, accurate in situ chemical and isotopic analyses of parts of single mineral grains are now possible, so that chemical evidence can be directly linked to individual mineral grains and their microstructural relationships. This combination of chemical and microstructural approaches is proving to be very powerful in the inference of rock-forming processes [e.g., Davidson et al., 1988, 2007; Charlier et al., 2005; Jerram and Davidson, 2007]. Just as limitations of whole-rock chemical analysis need to be acknowledged, so do limitations of structural interpretation, before applying it to specific problems.