Activation of the erector spinae during squat lifts depends on the initial posture of the lumbar spine. The authors assessed erector spinae activation by electromyography during squat lifts from lordotic and kyphotic postures, measured kinematics of the lifts from digitized video images, and inferred torques from the kinematics, using a two-dimensional model of a human lifting in the sagittal plane, with a joint at L3. Lifts from the lordotic initial posture had peak electromyographic signals early in the lift, whereas lifts from kyphotic initial posture had an initial "flexor relaxation," and peak activity in the middle of the lift. Lumbar flexion was much greater in lifts from kyphotic initial position. Torques required about L3 were similar between the two postures, though somewhat larger initially in lifts from kyphosis. The largest torques were therefore sustained by flexed lumbar spines, during periods of little or no erector spinae activity, in lifts made from kyphotic initial position. A sizable portion of the early torque is inertial, and therefore strongly dependent on movement time. Movements with a 30-lb load in the hands were similar, in kinematics and electromyography, to unloaded lifts, though longer in duration. The clinical implications of the differences in activation with posture, the practical implications of the inertial component of torque, and the need for consideration of lumbar posture in future modeling of squat lifting are discussed.