Recent researches have demonstrated the applicability of using Global Positioning System (GPS) techniques to precisely determine the 3-D coordinates of moving points in the field of natural hazards. Indeed, the detailed analysis of the motion of a landslide, in particular for a near real-time warning system, requires the combination of accurate positioning in three dimensions (infracentimetric) and fine temporal resolution (hourly or less). The monitoring of landslides with the GPS is usually performed using repeated campaigns, as a complement to conventional geodetic methods. Continuous monitoring of landslides with GPS is usually not performed operationally, mostly because of the cost of such a system compared to conventional deformation monitoring techniques. In addition, if GPS measurements can reach a millimetre-level accuracy for long observation sessions (typically 24 h), their accuracy decreases with the duration of the observation sessions, because of errors introduced by variations of the satellite constellation and multipath effects at the sites. This study aims at determining the experimental accuracy of GPS measurements for the continuous monitoring of landslides with GPS. In particular, we want to calibrate the variation of the measurement accuracy as a function of the duration of the observation sessions. The study was carried out on the Super-Sauze earthflow (Southern Alps, France) which evolves in a channelized flow with surface displacements reaching a few tens of centimetres to a few metres per year. The GPS data were acquired during two campaigns in May and October 1999 (two reference stations were installed outside the flow and four “moving” stations distributed on the flow). The maximal 3-D cumulative displacement reaches 2.1 m during 3 weeks in May 1999. The accuracy for a 1-h session reaches 2.7, 2.2 and 5.0 mm for the north–south, east–west and vertical components, respectively. The detectability threshold for a significant motion and a given temporal resolution stands between 3.5 mm/24 h and 8.5 mm/h in planimetry, between 6 mm/24 h and 19.5 mm/h in altimetry. Thus, the motion of the flow is clearly detected by the GPS measurements and the results have been compared with those obtained with conventional geodetic methods (theodolite and electronic distance-meters) or with a wire extensometer device. In addition, combination of periodical topometric measurements, continuous extensometric and GPS measurements allows us to identify seasonal and episodic transient variations in the surficial velocity of the flow. The analysis of the relationships between rainfall (and snowfall), groundwater level, and displacements permits us to understand the behaviour of the flow and to determine pore water pressures (PWP) thresholds initiating an acceleration of the movement. GPS therefore appears applicable to the continuous monitoring of geophysical objects or of man-made structures with small and slow displacements (∼5 mm/day). This technique does not require direct line of sight between the “moving” sites and the reference stations. Measurements can be carried out in all weather and at night. GPS processing can be performed in near real time without loss of accuracy. The use of GPS is, however, limited by the environmental characteristics of the geophysical object (mountains, vegetation), which can constitute masks limiting the visibility of the sky and create multipaths effects.