Measurements of landslide movement made by global positioning system surveys and extensometers over a 3.5-year period show that the Slumgullion landslide in the San Juan Mountains of southwest Colorado moved throughout the monitoring period, but that daily velocities varied on a seasonal basis. Landslide velocities peaked in the early spring and summer in response to snowmelt and summer thunderstorms, respectively. Velocities were slowest in mid-winter when air and soil temperatures were coldest and precipitation was generally low and/or in the form of snow with a low water content. We hypothesize that the seasonal variability in velocities is due to ground-water levels and corresponding pore pressures that decrease when surface water is unavailable or cannot infiltrate frozen landslide material, and increase when surface water from melting snow or rainfall infiltrates unfrozen landslide material. We also suggest that patches of bouldery debris and fractures (created by continuous movement of the landslide) are conduits through which surface water can infiltrate, regardless of the frozen or unfrozen state of the landslide matrix material. Therefore, the availability of surface water is more important than landslide temperature in controlling the rate of landslide movement. This hypothesis is supported by field instrumentation data that show (1) landslide velocities coinciding with precipitation amounts regardless of the depth of freezing of landslide material, (2) spring and annual landslide velocities that were greatest when the depth of freezing was also the greatest, and (3) a rapid (several weeks or less) velocity and pore pressure response to rainfall. The persistent, but seasonally variable movement of the landslide, fits the bathtub model for landslide movement described by Baum and Reid [Baum, R.L., Reid, M.E., 2000. Ground water isolation by low-permeability clays in landslide shear zones. In: Bromhead, E.N., Dixon, N., Ibsen, M.-L. (Eds.), Landslides in Research, Theory and Practice. Proc. 8th Int. Symp. on Landslides, Cardiff, Wales, vol. 1, 139-144]. In their model, the landslide is isolated both mechanically and hydrologically from adjacent materials by low permeability clays. These clays cause the landslide to retain water. Our data support this model by suggesting that pore pressures at the basal landslide surface are always adequate to maintain landslide movement and that any infiltration of water at the surface of the landslide is adequate to rapidly increase landslide velocity. ?? 2002 Elsevier Science B.V. All rights reserved.