The force that enlarges the cerebral ventricles and deforms the brain in hydrocephalus remains unclear. It is still widely thought to be elevated intraventricular pressure developing behind an obstruction to the flow of CSF. This view has led to the prediction that a large pressure difference should exist between the ventricles proximal to the obstruction and the subarachnoid space of the cerebral convexity distal to the obstruction. Yet measurements have shown consistently that such transmantle pressure differences are either small or absent. We propose a theory that reconciles the view that hydrocephalus is caused by obstruction to the flow of CSF with the observed absence of large pressure gradients across the cerebral mantle. Obstruction to CSF flow produces only a small pressure gradient -- usually less than 1 mm Hg -- that is sufficient to overcome the added resistance to flow and thereby to balance the absorption of CSF with its production. This mini-gradient is the effective force that initiates and sustains ventricular enlargement. It can coexist either with high or with normal intracranial pressure. The level of intracranial pressure is determined by the efficiency with which increments of ventricular pressure are transmitted through the parenchyma to the outer surface of the brain. In the presence of a rigid skull some transmission is required by basic laws of Newtonian mechanics. The efficiency of transmission depends primarily on the elastic properties of the brain. If the brain is relatively incompressible, transmission is efficient and high intracranial pressure is required to maintain the mini-gradient between the ventricles and the subarachnoid space, resulting in tension hydrocephalus. If the brain is more compressible, the parenchyma attenuates any increase of intraventricular pressure, reducing transmission to the outer surface. Intracranial pressure need not rise above normal levels to maintain the mini-gradient, leading to normal pressure hydrocephalus. The theory explains why tests measuring CSF resistance have limited diagnostic usefulness in hydrocephalus. It also predicts that very small stresses are sufficient to produce large deformations of the brain if these are allowed to occur slowly.