We report here the first attempt to map directly the boundary of a Vine-Matthews magnetic stripe on the seafloor. Our objectives are to study the processes of oceanic crystal accretion as recorded in the reversal trnasition zone and to investigate the formation of the magnetic source of Vine-Matthews magnetic anomalies. Our deep-tow and ALVIN-based magnetic studies focus on the Matuyama/Brunhes reversal transition on the flanks of the East Pacific Rise near 21 ¿N. While the sea level magnetic anomalies are less than average in clarity for the East Pacific Rise, a 'three-dimensional' inversion of deep-tow data reveals a sharp, strike-linear polarity transition less than 1.8 km wide (Macdonald et al., 1980a). These measurements have been augmented by mounting a vertical magnetic gradiometer on ALVIN and making 280 reliable polarity determinations along and across the polarity transition zone. Even on long traverses across both sides of the boundary, we find that nearly every magnetic target has the correct polarity, i.e., the same polarity as the regional magnetic lineation. This homogeneity in polarity of the magnetic lineations is surprising. The magnetic polarity transition in the outcropping volcanic section is sharp and linear along strike, delineated in some cases by a clear geologic contact of opposing flow fronts of different ages. Several weakly magnetized outcrops mapped within the transition zone may have erupted during the time in which the geomagnetic field was reversing. The reversal boundary mapped on the seafloor from ALVIN is displaced 250 m to 500 m NW away from the spreading axis relative to the position of the average boundary as derived from inversion of deep-tow and sea level magnetic data. This offset provides a means for estimating the spillover of lava flows away from the spreading axis during the time the crust was formed. The combination of deep-tow and ALVIN measurements suggest that circa 0.7 m.y. ago the crustal accretion zone (magnetized volcanic, intrusive, and plutonic rocks) was 2000--2800 m wide, while the zone of recent volcanism alone was only 1000--2000 m wide. The determination of the most recent reversal agrees well with submersible observations at the present spreading center where the zone of recent volcanism (neovolcanic zone) varies between 600 and 2000 m in width. This very orderly picture for the formation of magnetic lineations and crustal accretion processes appears to conflict with complex Deep Sea Drilling Project (DSDP) magnetic results from the Atlantic. We suggest that the crustal generating processes and resulting magnetic strcuture vary significantly with spreading rate. On the slow-spreading Mid-Atlantic Ridge, major episodes of volcanism are likely to be infrequent (~104 years), the magma chamber may be non-steady state, and the neovolcanic zone shifts or varies in width considerably. This sporadic, start and stop spreading process will contribute to a highly heterogeneous and complex crustal and magnetic structure as seen in DSDP holes. In addition, significant faulting and tilting may disrupt slow-spreading crust. For intermediate- to fast-spreading centers, more frequent volcanism (~50--600 years), a nearly steady state magma chamber, and a narrow, stable neovolcanic zone will create a less complex magnetic and crustal structure both as seen from ALVIN and as inferred from clear sea level magnetic anomalies in the Pacific.