Continuous permafrost 300 to 600 m thick is mapped in West Siberia beyond the Arctic Circle. To the south, it splits into two uneven layers to form a swallow-tail structure in cross-section. The surficial layer, which is about 50 m thick, occurs discontinuously between 66 ° and 64°N (in isolated patches up to the 60th parallel) and is thought to be a product of the Late Holocene cooling. The second, much thicker layer, separated from the surficial permafrost by a 100-200m thick talik, extends to the 60th parallel and must be a Pleistocene relict. Its fossil nature appears in near-zero temperatures and non-gradient curves of the vertical temperature profile (Yershov, 1989). The relict layer is commonly thought to be of the Late Weichselian age when solid permafrost reached the 51st parallel, i.e. shifted by 1500-1600 km south from its present limit (Baulin et al., 1984). Only Zemtsov (1976) related it to the Middle Pleistocene maximum glaciation. The study of fossil glacial ice contained in the thick discontinuous permafrost near the Arctic Circle has shown that relict permafrost must have survived since Early Weicheslian time (Astakhov and Isayeva, 1988). There are different ideas about the process which changed the ubiquitous solid permafrost of the Pleistocene into the present discontinuous swallowtail structure south of the Arctic Circle. The usual reasoning based on temperature profiles, freeze-andthaw structures in sediments and computer simulations does not allow unambiguous conclusions, particularly if geological data are disregarded. Balobayev et al. (1983) have suggested that the abrupt southern boundary of the modern continuous permafrost reflects a shoreline of a Late Weichselian proglacial lake that fringed the ice sheet along the Arctic Circle. Sedimentological research was discovered no traces of large proglacial lakes (Astakhov, 1989, 1992). Most investigators believe that the deep talik over the relict permafrost developed during the Holocene climatic optimum. The popular view, ascribing great mobility to the West Siberian permafrost, suggests that south of the Arctic Circle a permafrost layer approximately 500--600 m thick could form and vanish during the Late Pleistocene-Holocene (Baulin et al., 1981). This apparently contradicts geological data about the conservative nature of Siberian permafrost. The above mentioned lower Weichselian fossil ice and fresh-looking (not decayed) frozen wood of infinite radiocarbon age has survived within the permafrost near the surface (Astakhov and Isayeva, 1988). These facts show that: (i) no deep climatic fluctuations affected the permafrost since the Early Weichselian except the Holocene warming; and (ii) the Holocene thawing of the Arctic permafrost tens of metres deep (Baulin et al., 1981) must have been local. The apparent discrepancy between the data on great inertia of the Pleistocene permafrost in the Arctic and deep taliks registered south of the Arctic Circle must be attributed to heterogeneity of the degradation process in the different geographical zones (Astakhov, 1990). The divergence of geological processes in the Arctic from central West Siberia can be judged from the comparison of the sedimentary records. Facies related to the stable Weichselian permafrost show an arid and very continental climate from the Kara Sea to the Kazakh steppes. They are thick loess sheets (changing laterally into sand dunes), which are frozen with syngenetic ice veins in the Arctic, and thawed with ice wedge casts in the centre and south of West Siberia. The only zonal difference appears in the volume of water-lain facies within this principally aeolian formation. In the Arctic, aqueous sediments are mostly comprised of glaciofluvial sands and glaciolacustrine varves underlying the loess and dunes. Within the uppermost aeolian, mostly sandy formation, waterlain sediments are scarce, especially during 22 to 15 ka BP (Astakhov, 1991). Weichselian sediments in thermokarst sink-holes have not been described, even along the major river valleys. The situation is somewhat different south of the 64th parallel where numerous lenses of sandy, limnic rhythmites with dates from infinite to 22 ka BP are encompassed by loess-like silts with palaeosols and ice-wedge casts. They are especially abundant along the transverse Ob valley and the crest of the Siberian Hills. Such rhythmites, up to 15 m thick ('Kolpashevo sand' by Arkhipov et al., 1980), have nothing to do with large ice-damned lakes, because they:.(i) occur at all hypsometric levels from 40 to 120 m; (ii) contain seams of loess, palaeosols, cryoturbated horizons and ice-wedge casts; (iii) make oval and doughnut-shaped hillocks of thermokarst inversion topography; and (iv) are never traced laterally for more than a few hundred metres (Astakhov, 1989, 1992). These features indicate that in central West Siberia, thermokarst was active in the Middle Weichselian, probably along