The present-day active tectonics of the western Alps are poorly known. Permanent GPS stations located in the French and Italian Alps are too recent to give any significant information on the strain-regime within the chain [e.g. Calais et al., 2000a; Caporali and Martin, 2000]. Similarly, the reiteration in 1998 of the 60 points of the "GPS Alpes" temporary network, previously installed and positioned in 1993, did not result in a clear image of the active deformations of this part of the Alpine Arc [Vigny et al., 2001]. Both permanent and "GPS Alpes" data show that the relative motion of most of the points located within, or on both sides of the chain, are probably slower than 5 mm/yr. Another possibility to investigate the present-day deformation of part of the Alps is to use historical triangulation data. In many parts of the French Alps, authors have remeasured historical networks of the French Institut Geographique National, using GPS, for geodynamical purposes [Jouanne et al., 1994; Martinod et al., 1996; Ferhat et al., 1998; Sue et al., 2000; Calais et al., 2000b; Jouanne et al., 2001]. Their comparison confirms that deformations in the French Alps occur slowly, at speeds smaller than 5 mm/yr. Some deformations, however, have been observed in different parts of the chain [Jouanne et al., 1994; Martinod et al., 1996; Sue et al., 2000; Calais et al., 2000b]. Typically, the precision of triangulation data is 10 (super -5) , which means that the motion between benchmarks whose relative distance is 10 km must reach 10 cm to be noticed. Given the age of the triangulation networks that are re-measured using GPS (generally around 50 years), this corresponds to relative velocities of 2 mm/yr, which is quite large in the context of the western Alps. For instance, Martinod et al.  calculate a shortening axis orientated N070 degrees for the southern part of the Belledonne Massif (External Crystalline Massifs), and evaluate the relative speeds to reach possibly 3-5 mm/yr, which is as large as the maximum relative speed between Apulia and Europe! These results are based, however, on the motion of only 3 benchmarks (GGA, REV and GSE) of the historical network. In order to confirm the existence of the rapid deformation noted in this previous paper, we measured in 1998 and 1999, using GPS, the position of 22 historical benchmarks located near the southern part of the Belledonne Massif, which is the area where Martinod et al.  observed their most significant deformations. Geodetic data: 22 geodetic sites were measured using GPS in 1998 and 1999. Measurements were done using bi-frequency Ashtech receivers, in at least two 6-hour sessions for half of the points. 6 of those sites had already been measured in 1993-1994. We also included in the compensation of the GPS data the measurements of 4 sites (BUF, GEN, MCR and NER) that had been done in 1993 and 1994. GPS data have been processed using the Winprism software, and we used the Geolab software to perform the compensation of the 1993-1994 data together with the 1998-99 data. We finally obtain a new position for 26 benchmarks of the "Savoie-Dauphine 1950" triangulation network. We also performed again the compensation of the old triangulation network. We included in the compensation, data concerning the points of the geodetic campaign from the 1st order to the 4th order geodetic points. We calculated the position of 186 stations, using 1174 angle measurements. We assumed the standard deviation of a direction observation to result both from centering and instrumental errors [e.g. Jouanne et al., 1994]. We adopted the following uncertainties: 20 mm for centering errors, 6.3 10 (super -4) grads for Wild T3, and 7.6 10 (super -4) grads for Wild T2 theodolites (values communicated by IGN). The relative accuracy of the coordinates determined in this compensation is approximately 10 (super -5) . Comparison between triangulation and GPS data: It is not possible to obtain displacements vectors comparing GPS measurements with old triangulation data. As a matter of fact, historical geodetic networks only contain precise angle measurements. Neither the size, nor the orientation of the old network can be accurately known. To evidence possible tectonic deformations comparing the two geodetic campaigns, we calculate the strain tensor for triangular elements formed by sets of three neighbouring points of the network. We calculate the eingenvalues epsilon 1 and epsilon 2 of the strain tensor and their azimuth (resp. theta 1 and theta 2 ). We present in table II the values of dgamma /dt = (depsilon 1 /dt-depsilon 2 /dt) and of theta 2 for 33 triangles formed by sets of the 26 historical points remeasured using GPS. Both dgamma /dt and theta 2 are independent of the size and orientation of the old triangulation network. They can therefore be evaluated with precision without any a priori hypothesis [e.g. Ferhat, 1997]. dgamma /dt is the difference between the maximum compressive and extensive strain rate.