The involution structure in the Tokachi area of Hokkaido develops in two clearly bounded layers (upper: loam; lower: pumice), and the deformation pattern is fairly regular, with wave length of less than 1.5m and amplitude of 0.3 to 1.0m. Bulk density and viscosity of loam and pumice beds during the deformation were estimated from laboratory measurements. The bulk density of saturated loam is larger than that of saturated pumice. Under the assumption that the deformation was caused by this reversed density stratification, the wave-length of involution can be calculated by RAMBERG'S model. The result of the calculation is about 1.8m, which well approximates the field data. The vertically asymmetrical form of the interface of loam and pumice beds can be explained by a kinematic viscosity contrast of the two layers, as implied by ANKETELL (1970). When the driving force is only gravity, a reduction in the strength of the bed material is needed. Thawing of highly frozen ground, combined with poor drainage (flat topography over an impermeable bed), could offer the most suitable conditions for (over-) saturation and strength reduction of the fine-grained material. The involution structure shows that the continuity of deformation was interrupted by a cessation in strength reduction. Such short-term and extreme strength reduction might have resulted from such accidental events as liquefaction of saturated and cohesionless sediments. At least in this type of deformation, both RAMBERG'S and ANKETELL'S schemes are valid, although the possibility of secondary modifications of features by other periglacial actions such as frost heave should not be neglected.