The 1946 Aleutian earthquake stands out among tsunamigenic events because it generated both very high run-up near the earthquake source region and a destructive trans-Pacific tsunami. We obtained new data on the distribution of its tsunami in the near field along south-facing coasts between Unimak Pass on the west and Sanak Island on the east by measuring the height of driftwood and beach materials that were deposited by the tsunami above the extreme storm tide level. Our data indicate that (1) the highest measured run-up, which is at the Scotch Cap lighthouse, was 42 m above tide level or about 37 m above present storm tide elevation; (2) run-up along the rugged coast from Scotch Cap for 12 km northwest to Sennett Point is 12-18 m, and for 30 km east of Scotch Cap to Cape Lutke it is 24-42 m; (3) run-up along the broad lowlands bordering Unimak Bight is 10-20 m, and in-undation is locally more than 2 km; (5) run-up diminishes to 8 m or less at the southeast corner of Unimak Island; (6) no evidence was found for run-up above present storm tides (about 4-5 m above MLLW) on the Ikatan Peninsula or areas along the coast to the west; and (7) run-up above storm tide level in the Sanak Island group is restricted to southwest-facing coasts of Sanak, Long, and Clifford Islands, where it is continuous and locally up to 24 m high. Generation of the tsunami by one or more major earthquake-triggered submarine landslides near the shelf edge south of Unimak Island seems to be the only viable mechanism to account for the data on wave arrival time, run-up heights, and distribution, as well as for unconfirmed anecdotal reports of local postquake increases in water depth and diminished bottom-fisheries productivity. A preliminary hydrodynamic simulation of the local tsunami propagation and run-up using a dipolar model of a possible landslide off Davidson Bank provides an acceptable fit to the characteristics of the distribution of local run-up, with a value at 34 m at the Scotch Cap lighthouse.