AbstractOn the backwall of Twynam East Cirque selected stones moving through snow creep were marked and their movements from 1963 to 1975 measured. Their annual movements varied markedly between 1963 and 1970 but thereafter little. The percentage of stones failing to move varied reciprocally with snow thickness over a snow course in a comparable site near Spencers Creek. The annual mean movement, ranging from 7·6 to 36·3 cm, correlated strongly with Spencers Creek snow thickness. Frequent occurrence of annual movements of intermediate value characterise the process.Snow surface maps with 2m contours for 1964, 1974 and 1975 permitted determination of snow slope at the stone positions and of snow thickness by subtraction of ground heights. A 2 m contour ground surface map was also the basis for hillslope angles at stone positions. Basal area and area transverse to snow movement were calculated from measurements of the three principal axes of the stones. The vertical angles and directions of their abrasion tracks were measured.Rank correlation of stone movements with hillslope and snow slope was usually not significant but with track angle was so in most years. Track angle and direction are controlled partly by hillslope and partly by local rock geometry. Movements correlated inversely with stone basal area on which frictional drag depends. Correlation with stone transverse area was weak and inverse; this may be due to snow tending to move round larger stones. In most years there was significant correlation with snow thickness and basal stress (calculated from snow thickness and snow slope). However, the amount of ‘explanation’ associated with all the independent variables was small and the experiments revealed an importance of surface roughness at various scales in controlling stone movement which was much greater than expected from direct observ