Filament wound curved composite parts which are autoclave cured tend to exhibit "spring-in," a permanent deformation due to residual stresses (Stover, 1993). Spring-in is the tendency of a circular part to reduce it's radius of curvature upon radial cutting. The causes of spring-in and the associated residual stresses must be understood to ensure dimensional accuracy and safety of the final cured part. The purpose of this paper is to isolate and quantify the principle causes of spring-in in filament wound unidirectional autoclave cured carbon fiber/epoxy composite hoops. This paper first explores the causes of spring-in as given in the literature: through thickness inhomogeneity (Radford, 1993), part compaction (Meink, 1998), and anisotropy with initial curvature (Kollar, 1992). Next, the development of the residual stress profile through the fabrication and cure process is given along with the governing equations for the spring-in phenomenon. Finally, an experimental procedure is developed to obtain the residual stress profile in a thin composite part. This methodology involves progressive cutting (i.e., stress relief) and associated strain gage monitoring. The strain gage data is then combined with a finite element (FEM) analysis to determine the through-thickness residual stress profile. From the residual stress profile experiments, it was found that the principal cause of spring-in is not material anisotropy as was previously thought. Instead, the principle cause of spring-in was found to be the gradual thermal expansion of the mandrel during cure, which causes a "tension-lag" in the curing part. In addition, the compaction of the part during cure was found to have a nominal contribution to total spring-in.