The failure of existing theories to account for the type of crystal structure of a given metal is emphasized. It is suggested that, particularly when the number of bonding electrons is high, the metallic bond has greater directional characteristics than are generally assumed, and that these can be related to the symmetries of known hybrid bonds. Hybrids involving both gerade and ungerade orbitals (e.g. sp3) are not centro-symmetrical, and give rise to electron clouds biased on one side of the atom, which are thus strongly directional and suitable for strong bonding. Hybrids derived from gerade orbitals are centro-symmetrical, so that the binding electron has an equal chance of finding itself on opposite sides of the atom. This bonding is weaker, but the number of ligands is doubled; thus the four tetrahedral directions of the sd3 hybrid can give bonding to eight neighbours, as with an atom in a body-centred cubic structure. These ideas are applied to the face-centred cubic, body-centred cubic and close-packed hexagonal structures of the transition metals, and it is shown that the hybrids suggested by the crystal structures can be correlated with the known electron characteristics of these metals. An explanation of the difference between the crystal structures of the elements of groups VIIA and VIIIA of the first and those of the later transition series is advanced, and the general tendency for transition elements to be more soluble in the body-centred cubic than in the close-packed hexagonal modification of titanium and zirconium is explained.