The transverse creep response of an 8-ply SiC (SCS-6)/Ti-6Al-4V composite was measured at 482 ЊC from 69 to 276 MPa. Creep samples with fibers exposed at the edges as well as specimens with fully embedded fibers were tested under stepped loading conditions with increasing load. The response of each sample geometry was compared with creep data from the unreinforced matrix ('neat' material). The samples with exposed fiber ends exhibited minimum creep rates that were higher than those of the neat material at all stresses, and the stress exponent was slightly larger than the neat material. The embedded fiber samples possessed minimum creep rates that were smaller than the neat material at low stresses (Ͻ115 MPa), but became equivalent to the exposed fiber data at the highest stress (276 MPa). The apparent stress exponent for the embedded fiber composite was significantly larger than the neat material. The exposed fiber test data were well represented by a standard Crossman analysis, where the fibers were considered to have completely debonded. A stress singularity in the interfacial region at the sample edge is responsible for this behavior. The Crossman model was modified to incorporate the effect of a finite interface strength (120 MPa), and this was used to describe the response of the samples with embedded fibers. A reasonable fit to this representation was obtained. However, the measured minimum creep rate at the lowest stress was significantly lower than that predicted by the Crossman analysis for fully bonded fibers.