Retrospective climate simulations using the Goddard Institute for Space Studies SI2000 global climate model were used to examine possible causes of the observed upward trend in the frequency of U. S. extreme precipitation events. Three experiments were performed for the period 1951--1997, all with identical external radiative forcing but with different ocean treatments. One experiment used the Q flux ocean model. The second experiment imposed observed time-varying sea surface temperatures (SSTs). The third experiment imposed observed time-varying SSTs for the tropical Pacific only and imposed climatological SSTs elsewhere. The Q flux ocean experiment did not reproduce the observed upward trend. Both experiments using observed time-varying SSTs produced an upward trend qualitatively similar to the observed trend for 7-day duration events, although agreement was not as good for 1-day duration events. However, the experiment with time-varying SSTs only in the tropical Pacific was found to be overly sensitive to the increasing frequency of El Ni¿o events. Changes in atmospheric water vapor content for the two imposed-SST experiments were consistent with what might be expected to accompany increases in extreme precipitation. These results suggest that the spatial and temporal variations of SSTs on a global scale may be one direct and proximate cause of the recent observed upward trend in U. S. extreme multiday-duration precipitation events.