By modem gravity meters it is possible to m_r~_~Jre gravity differences with an accuracy higher than + 0,01 mgal ; even  0.001 regal and better becomes feasible under special conditions. Thus, more reliable information about the vertical gravity gradient is attainable by portable and relatively simple instrumentation. This fact is all the more important since the use of towers (like, e.g., trigonometric beacons or towers) has proven to be inefficient according to our experience ; for such measurements are mostly perturbed by wind effects. Using two steel frames (2m and 3m, high, respectively) with movable platforms an accuracy of the order of 10 -3 becomes feasible. In our experiments a model G LaCoste and Romberg.gravimeter was used which has a capcitance-readout--system with a strip chart recorder. In order to increase the accuracy and to get information about nonlinearity of dg/dh gravity was recorded at elevations h = 0; h = 0.5m; h = lm; h = 1.5m .... h=3mabovetheground. The error which is due to the imperfect knowledge of the real tidal perturbations can be avoided and hysteresis effects are reduced by recording gravity at avery elevation for about half an hour. It has been found that within this time elastic aftereffects and hysteresis had decreased so corresponding errors could be reduced. On this way we get from the record pieces of the ~idal curve at the point of observation which are shifted by translation with cesTpect to each other. Moreover, the microseismic effects are mostly averaged out by this procedure. In our c~__. the deflection of the gravimeter beam was recorded. As the calibration factor depends on the position of the beam a careful calibration of the gravimeter with respect to different reading lines is n~ry whenever ultimate accuracy is the goal: In (Groten, 1969) a corresponding calibration apparatus has been described .where heavy balls are used in order to produce gravitational differences in a laboratory. However, by using it we found that the temperature control of our gravity meter did not work so properly as we anticipated. Therefore, the dependence of the calibration function from the beam position has not yet been reliably established. But this is no principal defect. Moreover, the mass was not large enough in order to ensure an accuracy of :1:10 -4 which corresponds to 0.1/J~jal. �9 -- Premnmd at the Meeting of the International Grwity Commission, Paris, SePtarnber 1970.