High-grade (granulite-facies) terranes are brought to the surface by a combination of uplift and erosion (exhumation). The reported mechanisms and durations of exhumation are variable and depend partly on the mode of formation of a given high-grade terrane. In this paper, we consider the case of granulite-facies conditions that are attained (i) in juvenile crust, in the roots of magmatic arcs (e.g., Kohistan, Fiordland), (ii) around deep-seated high-temperature plutonic complexes, and (iii) in the lower parts of thickened continental crust. In the case of the roots of magmatic arcs, Phanerozoic examples suggest that they are exhumed along shallow-dipping contraction faults or shear zones that developed during continental obduction in a convergent tectonic regime. This process is not fundamentally different from processes leading to the exhumation of high-pressure (blueschist, eclogite) terranes. In contrast, deep-seated high-temperature plutonic complexes are thermostructural domes, analogous to the lower levels of core complexes, which may also have contributed to the uprise of high-grade terranes. Such domes should be sought for around anorthositic or mafic plutons, where their ascent may also have been favoured by continental extension. These modes of exhumation are compatible with a monocyclic evolution. However, many high-grade terranes show evidence of a polycyclic evolution and, in such cases, the nature of the thermal perturbation responsible for granulite-facies metamorphism is still debated. Thermal modelling based on heat conduction in collision orogens shows that granulites cannot form at mid-cristal levels, namely those exposed after isostatically driven denudation. Thus, magmatic underplating and crustal extension have been suggested as causes of steepened geotherms. Underplating (or intraplating) supplies the heat and thickens the crust from below. Postcollisional extension has also been considered as a mechanism providing a heat pulse emanating from the asthenosphere, probably after the "detachment" of a relatively cold thermal boundary layer. Finally, isolated crustal-scale intracratonic thrusting may favour the rise of intermediate to lower crustal wedges (e.g., the Kapuskasing wedge, uplifted prior to the trans-Hudson collision).