A series of field trials have been conducted for 11 months on and around Mt. Etna (Sicily) to test the performance of the new automated Scintrex CG-3M gravimeter on active volcanoes. The results show that the data from the CG-3M compare favourably with those from the traditionally employed LaCoste and Romberg (L&R) model D gravimeter. Under severe field conditions the Scintrex meter, because of its greater ruggedness and fully automated capabilities, yields results superior to those from the L&R D models. Repeated measurements (568) in a wide range of field conditions yielded errors in Δ g between stations of ± 3.5 to ± 4.4 μGal. Our tests also showed that both the strong drift in stationary conditions (~ 1 mGald −1 ) and changes of the instrumental calibration factor with time (whose rate decreased from 6 to 1 ppmd −1 during the 11-month study) can be reduced by regular checks to acceptable standards. Thus, (a) using experimental data, the instrumental drift was forecasted and corrected (by the real-time automatical compensator of the instrument) to within about 20 μGald −1 , and (b) using the measurements along an 18-station profile on Mt. Etna (total gravity range ≈ 300 mGal), and an 80-km-long calibration line away from the volcano (six stations, total gravity range = 365 mGal) the changes of the calibration factor were determined and corrected to within about 30 ppm.