Approximately 2000 arcs of Geos 3 altimeter data have been analyzed to determine geoid undulations and mean and point gravity anomalies. Since the altimeter data are subject to orbit error and bias, an error model was developed to remove such errors by comparing the altimeter geoid to the geoid implied by the Gem 9 coefficients. Using a one- or two-parameter error model and introducing crossover constraints, one primary adjustment involving 854 arcs distributed globally and six regional adjustments were performed. Typical a priori crossover discrepancies were ¿8 m, while a posteriori discrepancies were of the order of ¿6 m. The adjusted (bias removed) altimeter geoid undulations were used to construct 2-m contour interval geoids in selected regions and to determine gravity anomalies using least squares collocation. Using the data available up to September 1977, 377 equal-area anomalies of 5¿ and 12,144 mean anomalies of 1¿¿1¿ had been determined by February 1978. The average predicted accuracy was ¿3 mGal for the 5¿ blocks and ¿7 mGal for the 1¿ blocks in areas where sufficient altimeter data were available. These accuracy estimates appear to be reasonable when comparisons are made with terrestrial ground truth. The altimeter undulations were also used to derive point anomaly variations over areas of geophysical interest such as trenches and seamounts. three-dimensional polyhedra. By using Fourier transform methods the