Bridge abutments are designed to withstand lateral pressures from thermal expansion and seismic forces. Current design curves have been seen to dangerously over- and under-estimate the peak passive resistance and corresponding deflection of abutment backfills. Similar studies on passive pressure have shown that passive resistance changes with different types of constructed backfills. The effects of changing the length to width ratio, or including MSE wingwalls determine passive force-deflection relationships. The purpose of this study is to determine the effects of the wall heights and of the MSE support on passive pressure and backfill failure, and to compare the field results with various predictive methods. To compare the effects of backfill geometries, three large-scale tests with dense compact sand were performed with abutment backfill heights of 3 ft (0.91 m), 5.5 ft (1.68 m), and 5.5 ft (1.68 m) confined with MSE wingwalls. Using an existing pile cap 11 ft (3.35 m) wide and 5.5 ft (1.68 m) high, width to height ratios for the abutment backfills were 3.7 for the 3ft test, and 2.0 for the 5.5ft and MSE tests. The failure surface for the unconfined backfills exhibited a 3D geometry with failure surfaces extending beyond the edge of the cap, increasing the "effective width", and producing a failure "bulb". In contrast, the constraint provided by the MSE wingwalls produced a more 2D failure geometry. The "effective width" of the failure surface increased as the width to height ratio decreased. In terms of total passive force, the unconfined 5.5ft wall provided about 6% more resistance than the 5.5ft MSE wall. However, in terms of passive force/width the MSE wall provided about 70% more resistance than the unconfined wall, which is more consistent with a plane strain, or 2D, failure geometry. In comparison with predicted forces, the MSE curve never seemed to fit, while the 3ft and 5.5ft curves were better represented with different methods. Even with optimizing between both the unconfined curves, the predicted Log Spiral peak passive forces were most accurate, within 12% of the measured peak resistances. The components of passive force between the unconfined tests suggest the passive force is influenced more by frictional resistance and less by the cohesion as the height of the backwall increases.