Coupled thermal-mechanical models are used to investigate interactions between metamorphism, deformation and exhumation in large convergent orogens, and the implications of coupling and feedback between these processes for observed structural and metamorphic styles. The models involve subduction of suborogenic mantle lithosphere, large amounts of convergence (≥ 450 km) at 1 cm yr−1, and a slope-dependent erosion rate. The model crust is layered with respect to thermal and rheological properties — the upper crust (0–20 km) follows a wet quartzite flow law, with heat production of 2.0 μW m−3, and the lower crust (20–35 km) follows a modified dry diabase flow law, with heat production of 0.75 μW m−3. After 45 Myr, the model orogens develop crustal thicknesses of the order of 60 km, with lower crustal temperatures in excess of 700 °C. In some models, an additional increment of weakening is introduced so that the effective viscosity decreases to 1019 Pa.s at 700 °C in the upper crust and 900 °C in the lower crust. In these models, a narrow zone of outward channel flow develops at the base of the weak upper crustal layer where T≥600 °C. The channel flow zone is characterised by a reversal in velocity direction on the pro-side of the system, and is driven by a depth-dependent pressure gradient that is facilitated by the development of a temperature-dependent low viscosity horizon in the mid-crust. Different exhumation styles produce contrasting effects on models with channel flow zones. Post-convergent crustal extension leads to thinning in the orogenic core and a corresponding zone of shortening and thrust-related exhumation on the flanks. Velocities in the pro-side channel flow zone are enhanced but the channel itself is not exhumed. In contrast, exhumation resulting from erosion that is focused on the pro-side flank of the plateau leads to ‘ductile extrusion’ of the channel flow zone. The exhumed channel displays apparent normal-sense offset at its upper boundary, reverse-sense offset at its lower boundary, and an ‘inverted’ metamorphic sequence across the zone. The different styles of exhumation produce contrasting peak grade profiles across the model surfaces. However, P–T–t paths in both cases are loops where Pmax precedes Tmax, typical of regional metamorphism; individual paths are not diagnostic of either the thickening or the exhumation mechanism. Possible natural examples of the channel flow zones produced in these models include the Main Central Thrust zone of the Himalayas and the Muskoka domain of the western Grenville orogen.