ABSTRACTIn June 1994, the U.S. Air Force implemented a yaw bias in the attitude control subsystem of most GPS satellites. By January 1995, all Block II and IIA satellites were biased. The yaw bias was implemented to fix a problem with the attitude control subsystem that prevented the yaw attitude of the satellite from being properly modeled during eclipse seasons.This paper describes the problem that was identified with the GPS attitude subsystem, and explains its effects on the modeling of the satellite yaw attitude and, in turn, on the modeling of the carrier phase and pseudorange observables. It provides details of a new model for the GPS yaw attitude that is suitable for biased satellites during eclipse seasons. Early results using GPS with the biased attitude control subsystem are also presented.