We use GPS measurements at 37 stations in Honduras and El Salvador to describe active deformation of the western end of the Caribbean Plate between the Motagua fault and Central American volcanic arc. All GPS sites located in eastern Honduras move with the Caribbean Plate, in accord with geologic evidence for an absence of neotectonic deformation in this region. Relative to the Caribbean Plate, the other stations in the study area move west to west–northwest at rates that increase gradually from 3.3 ± 0.6 mm yr−1 in central Honduras to 4.1 ± 0.6 mm yr−1 in western Honduras to as high as 11–12 mm yr−1 in southern Guatemala. The site motions are consistent with slow westward extension that has been inferred by previous authors from the north-striking grabens and earthquake focal mechanisms in this region. We examine the factors that influence the regional deformation by comparing the new GPS velocity field to velocity fields predicted by finite element models (FEMs) that incorporate the regional plate boundary faults and known plate motions. Our modelling suggests that the obliquely convergent (∼20°) direction of Caribbean–North American Plate motion relative to the Motagua fault west of 90°W impedes the ENE-directed motion of the Caribbean Plate in southern Guatemala, giving rise to extension in southern Guatemala and western Honduras. The FEM predictions agree even better with the measured velocities if the plate motion west of the Central American volcanic arc is forced to occur over a broad zone rather than along a single throughgoing plate boundary fault. Our analysis confirms key predictions of a previous numerical model for deformation in this region, and also indicates that the curvature of the Motagua fault causes significant along-strike changes in the orientations of the principal strain-rate axes in the fault borderlands, in accord with earthquake focal mechanisms and conclusions reached in a recent synthesis of the structural and morphologic data from Honduras. Poor fits of our preferred models to the velocities of GPS sites near the Gulf of Fonseca may be an artefact of the still-short GPS time-series in this region or the simplifying assumptions of our FEMs.