We use broadband body and mantle wave data to study the 1998 Antarctic intraplate earthquake. The centroid moment tensor (CMT) has a large non-double-couple component. There exist two pure double-couple constrained solutions that fit the data almost equally well. The frequent practice of taking the best double-couple gives a far from optimal solution. We use P and SH body waves to determine the rupture parameters of the first and larger of the two observed subevents. The best rupture plane, with strike 96¿, dip 69¿, and rake -18¿, is compatible with only one of the two CMT solutions: strike 96¿, dip 64¿, rake -23¿, centroid location (63.1 ¿S, 148.4 ¿E, 10 km depth), centroid time 0313:02 UT, and M0=1.3¿1021 N m (Mw=8.0). The first subevent is a simple, primarily westward propagating ~140-km rupture, of ~45-s duration, with average velocity ≳3 km s-1; it has a seismic moment of 1.2¿1021 N m (Mw=8.0), with 75% of its moment released between 10 and 27 s, and a stress drop of ~240 bars. The rupture is physically bounded by two fracture zones at 147.5 ¿E and 150 ¿E. The second subevent lasted from 70 to 90 s on a fault extending from 210 to 270 km west of the epicenter, with a moment of 0.3--0.6¿1021 N m (Mw=7.6--7.8). This is a spectacular example of dynamic stress triggering over a 100-km separation distance with a time delay of ~40 s. The complex pattern of aftershocks is primarily controlled by preexisting fracture zones on the ocean floor.