The Yellowstone caldera, like many other later Quaternary calderas of the world, exhibits dramatic unrest. Between 1923 and 1985, the center of the Yellowstone caldera rose nearly one meter along an axis between its two resurgent domes (Pelton and Smith, 1979, Dzurisin and Yamashita, 1987). From 1985 until 1995-6, it subsided at about two cm/yr (Dzurisin and others, 1990). More recent radar interferometry studies show renewed inflation of the northeastern resurgent dome between 1995 and 1996; this inflation migrated to the southwestern resurgent dome from 1996 to 1997 (Wicks and others, 1998). We extend this record back in time using dated geomorphic evidence of postglacial Yellowstone Lake shorelines around the northern shore, and Yellowstone River levels in the outlet area. We date these shorelines using carbon isotopic and archeological methods. Following Meyer and Locke (1986) and Locke and Meyer (1994), we identify the modern shoreline as S1 (1.9 ? 0.3 m above the lake gage datum), map paleoshoreline terraces S2 to S6, and infer that the prominent shorelines were cut during intracaldera uplift episodes that produced rising water levels. Doming along the caldera axis reduces the gradient of the Yellowstone River from Le Hardys Rapids to the Yellowstone Lake outlet and ultimately causes an increase in lake level. The 1923-1985 doming is part of a longer uplift episode that has reduced the Yellowstone River gradient to a ?pool? with a drop of only 0.25 m over most of this 5 km reach. We also present new evidence that doming has caused submergence of some Holocene lake and river levels. Shoreline S5 is about 14 m above datum and estimated to be ~12.6 ka, because it post-dates a large hydrothermal explosion deposit from the Mary Bay area (MB-II) that occurred ~13 ka. S4 formed about 8 m above datum ~10.7 ka as dated by archeology and 14C, and was accompanied by offset on the Fishing Bridge fault. About 9.7 ka, the Yellowstone River eroded the ?S-meander?, followed by a ~5 m rise in lake level to S2. The lowest generally recognizable shoreline is S2. It is ~5 m above datum (3 m above S1) and is ~8 ka, as dated on both sides of the outlet. Yellowstone Lake and the river near Fishing Bridge were 5-6 m below their present level about 3-4 ka, as indicated by 14C ages from submerged beach deposits, drowned valleys, and submerged Yellowstone River gravels. Thus, the lake in the outlet region has been below or near its present level for about half the time since a 1 km-thick icecap melted from the Yellowstone Lake basin about 16 ka. The amplitude of two rises in lake and river level can be estimated based on the altitude of Le Hardys Rapids, indicators of former lake and river levels, and reconstruction of the river gradient from the outlet to Le Hardys Rapids. Both between ~9.5 ka and ~8.5 ka, and after ~3 ka, Le Hardys Rapids (LHR) was uplifted about 8 meters above the outlet, suggesting a cyclic deformation process. Older possible rises in lake level are suggested by locations where the ~10.7 ka S4 truncates older shorelines, and valleys truncated by the ~12.6 ka S5 shoreline. Using these controls, a plot of lake level through time shows 5-7 millennial-scale oscillations since 14.5 ka. Major cycles of inflation and deflation are thousands of years long. Le Hardys Rapids has twice been uplifted ~8 m relative to the lake outlet. These two locations span only the central 25% of the historic caldera doming, so that if we use historic doming as a model, total projected uplift would be ~32 m. This ?heavy breathing? of the central part of the Yellowstone caldera may reflect a combination of several possible processes: magmatic inflation, tectonic stretching and deflation, and hydrothermal fluid sealing and inflation followed by cracking of the seal, pressure release, and deflation. Over the entire postglacial period, subsidence has balanced or slightly exceeded uplift as shown by older shorelines that descend towards the caldera axis. We