The duration of ground acceleration, an important engineering seismology parameter, was determined for a subset of nuclear explosion and earthquake accelerograms. The nuclear-explosion, ground-motion data sample consisted of accelerograms derived from velocity recordings of high-yield events, MILROW and CANNIKIN, which have been assigned surface-wave magnitudes of 5.3 and 5.7, respectively. The earthquake-data sample consisted primarily of the Richter magnitude 6.6 San Fernando earthquake accelerograms. The criterion used to define duration of ground acceleration was the amount of time that the absolute acceleration is ≧ 5 per cent g, an approximate index of the strong phase of ground shaking. On the basis of this criterion and the data subset used, the duration of earthquake ground acceleration differs from that of nuclear explosions. The difference, which is not as great as has generally been thought in the past, is small inside 20 km. Earthquake durations are greater beyond 20 km. The scatter in the duration data was analyzed and found to be sensitive to local site amplification phenomena. These preliminary conclusions need to be validated by further research.