We present results of full-scale physical modelling of solifluction in two thermally defined environments: (a) seasonal frost penetration but no permafrost, and (b) a seasonally thawed active layer above cold permafrost. Modelling was undertaken at the Laboratoire M2C, Universite de Caen-Basse Normandie, Centre National de la Recherche Scientifique, France. Two geometrically similar slope models were constructed using natural frost-susceptible test soil. In Model I water was supplied via a basal sand layer during freezing. In Model 2 the basal sand layer contained refrigerated copper tubing that maintained a permafrost table. Soil freezing was from the top down in Model I (one-sided freezing) but from the top down and bottom up (two-sided freezing) in Model 2. Thawing occurred from the top down as a result of positive air temperatures. Ice segregation in Model I decreased with depth, but in Model 2, simulated rainfall led to summer frost heave associated with ice segregation at the permafrost table, and subsequent two-sided freezing increased basal ice contents further. Thaw consolidation in Model I decreased with depth, but in Model 2 was greatest in the ice-rich basal layer. Soil shear strain occurred during thaw consolidation and was accompanied by raised pore water pressures. Displacement profiles showed decreasing movement rates with depth in Model I (one-sided freezing) but 'phig-like' displacements of the active layer over a shearing basal zone in Model 2 (two-sided active layer freezing). Volumetric transport rates were approximately 2.8 times higher for a given rate of surface movement in the permafrost model compared with the non-permafrost model. Copyright (C) 2008 John Wiley & Sons, Ltd.