The incidence of plate-boundary earthquakes across 3 prospective tectonic segments at the Alaska subduction zone (ASZ) in the late Holocene is reconstructed from geological evidence of abrupt land-level change and archaeological evidence of discontinuities in occupation of native villages. Bracketing radiocarbon ages on uplifted and down-dropped coastal deposits indicate that great earthquakes likely ruptured the plate interface in the eastern segment (Prince William Sound [PWS]) about 800, 1400, 2200–2300, 2600–2700, 3100–3200, and 3600–3700 cal BP. Evidence for an event about 1900 yr ago, and the possibility that the 2600–2700 cal BP event was a closely spaced series of 3 earthquakes, is restricted to parts of Cook Inlet. Geological evidence from the central (Kenai [KEN]) segment is fragmentary, but indicates that this segment likely ruptured about 1400 yr ago and in the triple event about 2600–2700 yr ago. The geological record from the Kodiak-Katmai (KOKA) segment at the western end of the ASZ has limited time-depth, with localized evidence for ruptures about 500, 1000, and 1300 yr ago. 14C ages and stratigraphic descriptions from 82 prehistoric villages and camps on the coast of the Gulf of Alaska reveal fluctuations in site activity that correlate with paleoseismic episodes. Hiatuses in site occupation occurred about 800, 1400, and 2200 yr ago in the PWS and KEN segments. The fragmentary older record from the KEN segment also reveals a hiatus about 2700 yr ago. The 2200–2300 and 2600–2700 cal BP events are also recorded in the KOKA segment, and the great earthquake at about 3200 cal BP may also be recorded there. This suggests that, although the PWS and KEN segments behave as a coherent unit of the Alaska megathrust, the KOKA segment is characterized by semi-independent behavior. At least 2, and perhaps as many as 4, of the last 7 prehistoric great earthquakes at this plate boundary did not propagate this far west.