The mechanism of failure and the effect of augmentation of single pedicle screws subjected to caudo-cephalad loads applied at the screw head were investigated. In each of 10 lumbar vertebrae, Steffee pedicle screws of appropriate diameter were placed in one pedicle. The cancellous bone of the contralateral pedicle was removed by curettage, a custom fabricated bushing was pressed by hand into the space, and a Steffee screw was inserted through the bushing end into the vertebral body. This arrangement was designed to transfer load directly from the pedicle screw to cortex, bypassing the cancellous bone of the pedicle, in a second experiment, a custom plate was mounted to the posterior surfaces of the articular facets allowing load transfer from the head of the screw directly to cortex. Caudo-cephalad loads were applied by a materials tester and toggle displacement (defined as the total caudo-cephalad movement or the screw under minimal load after loading through a complete cycle) was measured. Under peak loads of +199 N (caudal) and −224 N (cephalad), mean toggle at the screw head was 4.93 mm (standard deviation [SD] = 3.60 mm) for the screw alone and 4.96 mm (SD - 4.42 mm) for the screw augmented by the bushing. Screws without augmentation showed a characteristic pattern of loosening with the base of the pedicle acting as a fulcrum and a butterfly-shaped void occurring in the vertebra and the pedicle where cancellous bone had been crushed. Problems with the bushing included poor fit, back-out, and split pedicles. By contrast, in the second group, the mean toggle displacement was 8.12 mm (SD = 3.06 mm) for the screw alone and 1.06 mm (SD = 2.49 mm) for the plate augmented screw (P = 0.0002). No failures were noted in the plate augmented screw group.