For more than a century, the origin and evolution of the set of cuspate forelands known as the Carolina Capes—Hatteras, Lookout, Fear, and Romain—off the eastern coast of the United States have been discussed and debated. The consensus conceptual model is not only that these capes existed through much or all of the Holocene transgression, but also that their number has not changed. Here we describe bathymetric, lithologic, seismic, and chronologic data that suggest another cape may have existed between Capes Hatteras and Lookout during the early to middle Holocene. This cape likely formed at the distal end of the Neuse-Tar-Pamlico fluvial system during the early Holocene transgression, when this portion of the shelf was flooded ca. 9 cal (calibrated) kyr B.P., and was probably abandoned by ca. 4 cal kyr B.P., when the shoreline attained its present general configuration. Previously proposed mechanisms for cape formation suggest that the large-scale, rhythmic pattern of the Carolina Capes arose from a hydrodynamic template or the preexisting geologic framework. Numerical modeling, however, suggests that the number and spacing of capes can be dynamic, and that a coast can self-organize in response to a high-angle-wave instability in shoreline shape. In shoreline evolution model simulations, smaller cuspate forelands are subsumed by larger neighbors over millennial time scales through a process of ‘cape capture.’ The suggested former cape in Raleigh Bay represents the first interpreted geological evidence of dynamic abandonment suggested by the self-organization hypothesis. Cape capture may be a widespread process in coastal environments with large-scale rhythmic shoreline features; its preservation in the sedimentary record will vary according to geologic setting, physical processes, and sea-level history.