Dual-frequency transmissions from the Global Positioning System satellites can be used to measure and map ionospheric total electron content (TEC) on global scales. Using data exclusively from ground-based GPS networks, global ionosphere mapping has been successfully applied using either two- or three-dimensional techniques. Two-dimensional TEC maps retrieve a horizontally-varying distribution of total electron content, assuming a fixed vertical electron density profile. In three-dimensional mapping, both the horizontal and vertical distribution density are adjusted to fit the data. We describe a three-dimensional TEC mapping algorithm that uses three independent constant-density slabs stacked vertically to model the electron density, and compare with a more conventional two-dimensional approach using a single slab. One apparent benefit of the new method is reduction in a level error of the TEC maps, which decreased by 1.7 TECU using the three-dimensional retrieval on simulated data (1 TEC Unit corresponds to 10¹⁶ electrons/m²). Another benefit of the multilayer approach is improved slant TEC modeling. Using actual data from an equatorial site at Cocos Islands (96.8E, 12.2 S), three slab modeling improved estimates of slant TEC by a factor of 2 for elevation angles between 10 and 20° (9 versus 4.4 TECU, root-mean-square). However, the global structure of the vertical TEC retrievals we analyzed did not improve using three-dimensional modeling. This may be due to a critical approximation shared by both techniques that TEC persists unchanged at a given local time. This assumption is required to produce global maps from observations acquired from widely scattered ground receivers. Further improving the retrieval of global TEC structure with ground-based data probably requires improved dynamical models of TEC behavior. New data available from GPS receivers in low Earth orbit is also promising.