The currently adopted approach to reduce observed gravity data for geophysical purposes includes several approximations. These were originally used to reduce computational effort, but have remained standard practice, even though the required computing power is now readily available. In contrast, more precise gravity reductions are routinely employed in physical geodesy. The difference between simple Bouguer gravity anomalies derived using the geophysical and geodetic approaches can reach several tens of mu m sec(-2). The geodetic reductions include a more accurate calculation of normal gravity as a function of latitude, and a free air correction that accounts for the non-sphericity of the figure of the Earth. Also important, especially given the advent of Global Positioning System coordination of gravity surveys, is the need to ensure that the correct vertical and horizontal coordinate systems are used for the gravity reduction procedure. Errors associated with tie use of non-geocentric horizontal coordinates and ellipsoidal heights are significant when compared with the accuracy of an individual gravity measurement. A generalised gravity reduction program and a coordinate transformation program are presented which can be employed to reduce geophysical data in a geodetic manner. (C) 1998 Elsevier Science Ltd.