We compare the timing and spatial extent of dipolarization with auroral breakup by using magnetic field data (1-min time resolution) from GOES 8 and GOES 9 and global auroral images (37-s frame rate) from the Polar ultraviolet imager. We survey the entire year of 1997 data and obtain 32 unambiguous dipolarization events that had ionospheric footprints within 2 hours in magnetic local time (MLT) from their associated auroral breakups. It is found that the auroral breakup preceded the arrival of dipolarization at GOES by an average of 1.7 ¿ 2.7 min. (1.0 ¿ 2.1 min for events ¿1 hour from the auroral breakup). The delay of geosynchronous dipolarization represents a propagation time, as the substorms rarely initiated at the GOES location. Propagation velocities derived from the propagation time are ~60 km s-1 eastward, ~75 km s-1 westward, and ~270 km s-1 earthward at geosynchronous orbit. We also show for the first time that auroral bulge and dipolarization region closely map; the bulge is the region of currents out of the ionosphere. The eastern edge of the dipolarization region, where field-aligned currents are flowing into the ionosphere, is, based on the T89 magnetic field model, located at ~0.5 hour in MLT eastward of the eastern edge of the auroral bulge. Within the initiation site meridian the dipolarization onset was observed ~1 min prior to the auroral breakup at the GOES altitude. This time is associated with Alfv¿n transit time and therefore depends on the radial location of the onset. The azimuthal scale of the initiation site is found to be ~0.5 hour in MLT in the ionosphere or less than ~1 Earth's radius at 6.6 RE. These results support current diversion into the ionosphere as a critical part of substorms. Finally, a simple timing analysis of specific substorm onset time features suggests that near-Earth reconnection cannot occur before current disruption and therefore cannot be the cause of substorm.