Satellite studies of magnetospheric substorms are necessarily statistical since there are few properly instrumented satellites in space at any one time. To carry out these studies, one selects an ensemble of similar substorms, determines an origin for substorm time, and performs an ensemble average to eliminate noise. Similarity between different substorms can be determined from contour maps of midlatitude magnetic perturbations as a function of longitude and universal time. The origin of substorm time can also be determined from midlatitude perturbations and is usually taken as the beginning of the positive bay near midnight. Use of these techniques has led to a phenomenological model of magnetospheric substorms that includes a growth phase. The controversy over the existence of the growth phase is at least partially a consequence of differing definitions of the origin of substorm time. Use of auroral zone magnetograms often gives an earlier onset than midlatitude magnetograms. However, midlatitude onsets do appear to order satellite data systematically. In this paper we illustrate the discrepancy between these two onset times and show how midlatitude onsets order satellite data in agreement with the growth phase model. Three events of increasing complexity are considered. Each has been previously reported in the literature and later criticized. We speculate that a polar magnetic substorm begins when a pair of X- and O-type neutral lines form on closed field lines within the plasma sheet. Subsequently, the midlatitude positive bay begins when magnetic merging first encounters open field lines in the lobe of the tail.