An important feature of natural ice, in addition to the obvious relevance of glaciers and ice sheets for climate-related issues, is its ability to creep on geological time scales and low deviatoric stresses at temperatures very close to its melting point, without losing its polycrystalline character. This fact, together with its strong mechanical anisotropy and other notable properties, makes natural ice an interesting model material for studying the high-temperature creep and recrystallization of rocks in Earth’s interior. After having reviewed the major contributions of deep ice coring to the research on natural ice microstructures in Part I of this work (Faria et al., 2014), here in Part II we present an up-todate view of the modern understanding of natural ice microstructures and the deformation processes that may produce them. In particular, we analyze a large body of evidence that reveals fundamental flaws in the widely accepted tripartite paradigm of polar ice microstructure (also known as the “three-stage model,” cf. Part I). These results prove that grain growth in ice sheets is dynamic, in the sense that it occurs during deformation and is markedly affected by the stored strain energy, as well as by air inclusions and other impurities. The strong plastic anisotropy of the ice lattice gives rise to high internal stresses and concentrated strain heterogeneities in the polycrystal, which demand large amounts of strain accommodation. From the microstructural analyses of ice cores, we conclude that the formation of many and diverse subgrain boundaries and the splitting of grains by rotation recrystallization are the most fundamental mechanisms of dynamic recovery and strain accommodation in polar ice. Additionally, in fine-grained, high-impurity ice layers (e.g. cloudy bands), strain may sometimes be accommodated by diffusional flow (at low temperatures and stresses) or microscopic grain boundary sliding via microshear (in anisotropic ice sheared at high temperatures). Grain boundaries bulged by migration recrystallization and subgrain boundaries are endemic and very frequent at almost all depths in ice sheets. Evidence of nucleation of new grains is also observed at various depths, provided that the local concentration of strain energy is high enough (which is not seldom the case). As a substitute for the tripartite paradigm, we propose a novel dynamic recrystallization diagram in the three-dimensional state space of strain rate, temperature, and mean grain size, which summarizes the various competing recrystallization processes that contribute to the evolution of the polar ice microstructure. Dedicated to Sepp Kipfstuhl on occasion of his 60th birthday.