The CASA UNO Global Positioning System (GPS) experiment (January--February 1988) included an extended tracking network which covered three continents in addition to the network of scientific interest in Central and South America. The repeatability of long baselines (400--1000 km) in South America is improved by up to a factor of two in the horizontal vector baseline components by using tracking stations in the Pacific and Europe to supplement stations in North America. In every case but one, the differences between the mean solutions obtained using different tracking networks was equal to or smaller than day-to-day rms repeatabilities for the same baselines. The mean solutions obtained by using tracking stations in North America and the Pacific agreed at the 2--3 millimeter level with those using tracking stations in North America and Europe. The agreement of the extended tracking network solutions suggests that a broad distribution of tracking stations provides better geometric constraints on the satellite orbits and that solutions are not sensitive to changes in tracking network configuration when an extended network is used. A comparison of the results from the North Andes and a baseline in North America suggests that the use of a geometrically strong extended tracking network is most important when the network of interest is far from North America.