Throughout the Early Tertiary the area of the Farallon oceanic plate was episodically diminished by detachment of large and small northern regions, which became independently moving plates and microplates. The nature and history of Farallon plate fragmentation has been inferred mainly from structural patterns on the western, Pacific-plate flank of the East Pacific Rise, because the fragmented eastern flank has been subducted. The final episode of plate fragmentation occurred at the beginning of the Miocene, when the Cocos plate was split off, leaving the much reduced Farallon plate to be renamed the Nazca plate, and initiating Cocos-Nazca spreading. Some Oligocene Farallon plate with rifted margins that are a direct record of this plate-splitting event has survived in the eastern tropical Pacific, most extensively off northern Peru and Ecuador. Small remnants of the conjugate northern rifted margin are exposed off Costa Rica, and perhaps south of Panama. Marine geophysical profiles (bathymetric, magnetic and seismic reflection) and multibeam sonar swaths across these rifted oceanic margins, combined with surveys of 30-20 Ma crust on the western rise-flank, indicate that (i) Localized lithospheric rupture to create a new plate boundary was preceded by plate stretching and fracturing in a belt several hundred km wide. Fissural volcanism along some of these fractures built volcanic ridges (e.g., Alvarado and Sarmiento Ridges) that are 1-2 km high and parallel to "absolute" Farallon plate motion; they closely resemble fissural ridges described from the young western flank of the present Pacific-Nazca rise. (ii) For 1-2 m.y. prior to final rupture of the Farallon plate, perhaps coinciding with the period of lithospheric stretching, the entire plate changed direction to a more easterly ("Nazca-like") course; after the split the northern (Cocos) part reverted to a northeasterly absolute motion. (iii) The plate-splitting fracture that became the site of initial Cocos-Nazca spreading was a linear feature that, at least through the 680 km of ruptured Oligocene lithosphere known to have avoided subduction, did not follow any pre-existing feature on the Farallon plate, e.g., a "fracture zone" trail of a transform fault. (iv) The margins of surviving parts of the plate-splitting fracture have narrow shoulders raised by uplift of unloaded footwalls, and partially buried by fissural volcanism. (v) Cocos-Nazca spreading began at 23 Ma; reports of older Cocos-Nazca crust in the eastern Panama Basin were based on misidentified magnetic anomalies. There is increased evidence that the driving force for the 23 Ma fission of the Farallon plate was the divergence of slab-pull stresses at the Middle America and South America subduction zones. The timing and location of the split may have been influenced by (i) the increasingly divergent northeast slab pull at the Middle America subduction zone, which lengthened and reoriented because of motion between the North America and Caribbean plates; (ii) the slightly earlier detachment of a northern part of the plate that had been entering the California subduction zone, contributing a less divergent plate-driving stress; and (iii) weakening of older parts of the plate by the Galapagos hotspot, which had come to underlie the equatorial region, midway between the risecrest and the two subduction zones, by the Late Oligocene.