Recent multi-site investigations along the 1959 Mw 7.3 Hebgen Lake earthquake rupture (Hebgen Lake Paleoseismology Working Group, 2000) found evidence of two late Holocene surface-rupturing earthquakes (including 1959) on the Hebgen and Red Canyon faults. Trench and topographic relations at the Section 31 site near the center of the Hebgen fault also indicated a possible third event. In the summer of 2001, we returned to Section 31 to deepen and extend previous trenches across a compound scarp formed on an alluvial fan estimated from cosmogenic surface dating to be 11-15 ka. The new exposures provided clear evidence of the third, or pre-penultimate, earthquake. This event appears to closely post-date the sequence of alluvial-fan deposition and may have been responsible for abandonment of the fan surface. The event's antiquity is evidenced by a gently sloping fault scarp (modeled as having a morphologic age of 6-14 ka) and an argillic soil horizon that developed on the third-event colluvial wedge and subjacent to the fault scarp. Subsequent faulting in the late Holocene stepped several meters into the hanging wall, isolating much of the oldest wedge and scarp on the up-thrown block. Third-event colluvial deposits capped by organic material that had accumulated at the toe of the degraded scarp were exposed in the free faces of the two younger scarps and became the primary source of material for the penultimate and 1959 colluvial wedges. Net tectonic vertical displacement of the alluvial-fan surface is 5-6 meters; net displacement in the past two events is 2-3 meters. Displacement of the modern drainage at the main fault indicates a minimum of ~1.2 meters of vertical slip at the site in 1959. The scarp associated with the penultimate earthquake has been obliterated at the site by upslope retreat of the 1959 scarp. Remnants of a steep slope above and below the 1959 scarp can be seen on photographs taken shortly after the earthquake. By 1978, when follow-up photography was taken at the site, the penultimate scarp had been largely eroded away. The rapid disappearance of the prior-event scarp cautions against assuming that a morphologically simple scarp or the bevel of a compound scarp represents a single event.