In areas that are seismically active but lacking clear surficial faulting, many paleoearthquake studies depend on the interpretation of ancient liquefaction features (sand blows) as indicators of prehistoric seismicity. Sand blows, however, can be mimicked by nonseismic sand boils formed by water seeping beneath levees during floods. We examined sand boils induced by the Mississippi River flood of 1993 in order to compare their characteristics with sand blows of the New Madrid earthquakes of 1811–1812. We found a number of criteria that allow a distinction between the two types of deposits. (1) Earthquake-induced liquefaction deposits are broadly distributed about an epicentral area, whereas flood-induced sand boils are limited to a narrow band along a river's levee. (2) The conduits of most earthquake-induced sand blows are planar dikes, whereas the conduits of flood-induced sand boils are most commonly tubular. (3) Depression of the preearthquake ground surface is usual for sand blows, not for sand boils. (4) Flood-induced sand boils tend to be better sorted and much finer than sand-blow deposits. (5) Source beds for earthquake-induced deposits occur at a wide range of depths, whereas the source bed for sand boils is always near surface. (6) Materials removed from the walls surrounding the vent of a sand blow are seen inside sand blows, but are rarely seen inside sand boils. In general, flood-induced sand boils examined are interpreted to represent a less-energetic genesis than earthquake-induced liquefaction.