Monte Carlo studies show that an exclusion principle, based upon a minimum allowed O‐O separation in adjacent tetrahedra, drastically restricts the relative orientations such tetrahedra can assume. This ad hoc criterion, with no energy considerations involved, reproduces the experimentally observed Si‐O‐Si angle distribution determined from the structural data of a large number of silicas and silicates. Similar results are obtained when Coulombic or nonbonded energy calculations are used. When extended structures are formed, this steric requirement alone causes additional and severe restrictions on the relative orientations of adjacent tetrahedra. In fact, it is not possible to have completely random orientations of tetrahedra beyond the first neighborhood shell, unless some nonrandomness is introduced to help formation of closed rings. Thus, the completely random continuous network model fails to produce extended structures in vitreous silica and defect models must be introduced to patch up the model. These considerations also apply to GeO2and SiS2structures.