The current state of understanding of recrystallization texture development and of recrystallization texture modeling and prediction is reviewed. It is demonstrated that classical approaches as well as current models can only explain measured textures but fail to reliably predict textures. The strengths and deficiencies of major developments are discussed. We demonstrate that the interdependency of a correct growth law and an appropriate nonrandom nucleus texture effectively controls the emergence or suppression of recrystallization texture components. Without deeper insight into these fundamental microstructural processes a quantitative texture prediction is bound to fail no matter how sophisticated the used approach. In essence, we do not need better recrystallization models, we need better models and data of nucleation and growth.