Various observations pointed out that cores performed with gravity piston corer show signifi cant distortions mainly located at the top of the core. A series of 15 cores were performed at the same location on a submarine sand wave (Var canyon, France). Six different settings of the corer – three freefall heights and three slacks of the piston cable – were tested, including duplicates. Two accelerometers recorded simultaneously the movements of the core tube and the movements of the triggering arm. Then the z displacements were obtained by a double integration versus time of the measured acceleration. The analyses of results allowed the authors to estimate the amplitude and the duration of the elastic recoil of the aramid cable, andto distinguish four steps during the 4 seconds of penetration, including a distortion phase followed by a normal sampling phase linked to the status of the piston. The analyses of the quality and benchmark layers from recovered cores highlight the major role of the piston driven by the lengths of the counterweight and piston cables. The recovered thickness of a given layer can vary from 0.8 to 1.3 depending to the settings. A cone penetrometer test (CPT) trial at the same location gives a good estimation of the absolute geometry of the layers. The settings for cores with geotechnical purpose (better quality) will be different from settings for cores with sedimentological or palaeoclimatological purposes (better geometry). A compromise is proposed.