Hummocks (nonsorted circles) are widely distributed in arctic and subarctic regions. The hummocks under discussion are composed of fine-grained frost-susceptible soils; the late summer frost table is bowl-shaped; and the hummocks grade from those which are completely vegetated (earth hummocks) to those with bare centres (mud hummocks). The mound form is usually attributed to an upward displacement of material resulting from cryostatic (freeze-back) pressure generated in a confined, wet, unfrozen pocket of the active layer. Theoretically, cryostatic pressure should not develop in a frost-susceptible hummock soil, because ice lensing at the top and (or) bottom of the active layer will desiccate the last unfrozen pocket so that the pore water is under tension, not pressure. Field observations carried out at Garry Island, Northwest Territories, for 1965–1979 and for 1967–1979 at Inuvik, Northwest Territories, involving: summer and winter excavations; the measurement of the deformation of tubes, soil pressure, soil temperature, soil heave, soil moisture migration; and observations on hummock stability provide no field evidence for the cryostatic theory. An alternative model of hummock growth, based upon field observations, is here proposed. The upward displacement of material is believed caused by freeze–thaw of ice lenses at the top and bottom of the active layer with a gravity-induced cell-like movement, because the top and bottom freeze–thaw zones have opposite curvatures. The cell circulation is evident from the grain-size distribution of hummock soils and from upward-moving tongues of saturated soil observable in late summer. The most active period is in late summer. Model experiments in the laboratory have been successful in producing mounds by freeze–thaw of a kaolin slurry in a bowl-shaped container in support of the proposed theory.