Gravity was measured at eight different heights on a 300-m meteorological tower, using electrostatically nulled LaCoste and Romberg gravimeters. The observed values were corrected for tides, temperature, and gravimeter screw errors, and tested for systematic effects due to (wind-induced) tower motion (no such effects were found). These corrected results were compared with values predicted by means of Newton's inverse square law from surface gravity values. The differences exhibit no systematic trends, and their rms value is only 10 μGal, well within the errors of the experiment, as the estimated measurement errors increase from 9 μGal at the lowest platform to 14 μGal at the top and those of the predictions from 10 μGal to 23 μGal. These results set new constraints on the magnitude of any non-Newtonian gravitational force; if such a force is derived from a Yukawa potential, the absolute value of α must be less than 0.001 for λ=1000 m. ¿American Geophysical Union 1991