Erosive features named "regressive alcoves" had been studied on na experimental field site, in which soil materials are predominantly non-cohesive. Frequently associated by the literature to seepage erosion, those erosive features proved to be mostly controlled by specific mechanisms associated to hortonian overland flow. During the experiment, solid particles transported by overland flow and by subsurface flow were collected separated, after natural rain events. After each event, the total transported load was weighed and erosion rates estimated. The morphologic evolution of the erosive feature was followed in the field by the application of topographic techniques, adapted to the studied problem. Most of the gully head expansion was associated to the retreat of the "regressive alcoves". Although subsurface flow became of some importance during specific events, most of the erosion measured at the erosive features was associated to overland flow. Regressive alcoves are very common erosive features on tropical areas, and elsewhere. As one of the first observations needed to gully erosion control surveys is the precise identification of the points in which specific agents are important to gully erosion, the commonly accepted view regarding those erosive features, according to which regressive alcoves evolve as a consequence of seepage erosion, may lead, at least, to an expensive and unnecessary control design.