Simultaneous measurements of vertical accretion and change in surface elevation relative to a shallow (3-5 m) subsurface datum were made in selected coastal salt marshes of Louisiana, Florida, and North Carolina to quantitatively test Kaye and Barghoorn's contention that vertical accretion is not a good surrogate for surface elevation change because of autocompaction of the substrate. Rates of subsidence of the upper 3-5 m of marsh substrate were calculated for each marsh as the difference between vertical accretion and elevation change measured with feldspar marker horizons and a sedimentation-erosion table. Surface elevation change was significantly lower than vertical accretion at each site after 2 years, indicating a significant amount of shallow subsidence had occurred, ranging from 0.45 to 4.90 cm. The highest rate of shallow subsidence occurred in the Mississippi delta. Results confirm Kaye and Barghoorn's contention that vertical accretion is not generally a good surrogate for elevation change because of processes occurring in the upper few meters of the substrate, including not only compaction but also apparently shrink-swell from water storage and/or plant production--decomposition at some sites. Indeed, surface elevation change was completely decoupled from vertical accretion at the Florida site. The assumption of a 1:1 relationship between accretionary and substrate processes. Consequently, the potential for coastal marsh submergence should be expressed as an elevation deficit based on direct measures of surface elevation change rather than accretion deficits. These findings also indicate the need for greater understanding of the influence of subsurface and small-scale hydrologic processes on marsh surface elevation.