In human standing, gravity causes forward toppling about the ankle joint which is prevented by activity in the soleus and gastrocnemius muscles. It has long been assumed that when people sway forwards the calf muscles are stretched and conversely that they shorten with backward sway. Consequently, for many years, two explanations for standing stabilisation have flourished. First, tonic muscle activity itself may generate adequate intrinsic ankle stiffness. Second, if intrinsic ankle stiffness is inadequate, the resistance to stretch of the calf muscles may be augmented by stretch reflexes or by central control. These explanations require that the passive tissue (Achilles tendon, foot) transmitting the calf muscle tension is stiff. However, our recent measurements have indicated that this passive tissue is not stiff during standing. Accordingly, we predicted a counterintuitive mode of control where the muscles and body must, on average, move in opposite directions (paradoxical movements). Here we use dynamic ultrasound imaging in vivo with novel automated tracking of muscle length to test our hypothesis. We show that soleus and gastrocnemius do indeed move paradoxically, shortening when the body sways forward and lengthening when the body returns. This confirms that intrinsic ankle stiffness is too low to stabilise human standing. Moreover, it shows that the increase in active tension is associated with muscle shortening. This pattern cannot be produced by muscle stretch reflexes and can only arise from the anticipatory neural control of muscle length that is necessary for balance.