The equilibrium structures of the complexes M(n+)L(m) (M(n+) = K+, Rb 1+, Cs+, Ca2+, Sr2+, Ba2+, Sc3+, Y3+, La3+; L = NH3, H2O, HF; m = 1-3) have been computed ab initio by using quasirelativistic pseudopotentials and flexible, polarized basis sets. Many of these species do not obey expectations based on the valence shell electron pair repulsion (VSEPR) rules or simple electrostatic models. For m = 2 (except for K+) bent L-M-L arrangements are favored energetically over linear structures. The energy gain upon bending increases along the series M(n+) = K+, Rb+, Ca2+, Cs+, Sr2+, Ba2+, Sc3+, Y3+, La3+. The smallest angles (ca. 110-degrees) and largest linearization energies (up to ca. 7 kcal/mol for La3+(NH3)2) are found with Ba2+, Sc3+, Y3+, and La3+. Complexes of Ba2+ and La3+ with three NH3, H2O, or HF ligands exhibit a preference for pyramidal over trigonal-planar arrangements although the pyramidalization energy is less than 1 kcal/mol. While the main reason for the very small bending effects in the group 1 cations is the polarization of the cation by the field of the ligands, the participation of d orbitals in covalent bonding contributions seems to be the major driving force for the group 3 cations. Both effects probably are important for the group 2 cations. The observed angles in the M(n+)L2 complexes are considerably smaller than those for the alkaline-earth metal dihalides, dihydroxides, or diamides; La3+(HF)3 is more pyramidal than LaF3. These smaller angles are due to decreased repulsion between neutral ligands as compared to anions and to reduced pi-bonding in the cationic complexes. Extended d-basis sets are needed for the computation of heavy alkaline-earth metal cations. In particular, model ab initio calculations used previously to parametrize semiempirical force fields for calcium protein simulations suffered from the use of inadequate Ca basis sets.