The force field for the cellular ligand field stabilisation energy/molecular mechanics (CLFSE/MM) method has been applied to 28 transition-metal complexes. Computed and experimental structures are compared for 12 ML _x Cl _{6- x} species (M = Co ^III or Ni ^II ; L = amine donor; x = 6 or 4), 12 MA _x B _{6- x} compounds (M = Ni ^II or Cu ^II ; A = imine, B = amine; x = 6, 4 or 3), one five-co-ordinate copper(II) imine–amine complex and three four-co-ordinate copper(II) imine and imine–amine molecules. For π-bonding ligands a stronger donor interaction is associated with a larger positive value of the CLF e _π parameter but, due to the use of a crystal field type barycentre, the CLFSE actually goes up. The CLFSE thus has the wrong form for treating the π contributions to bond stretching and distance-dependent e _π parameters are inappropriate. However, the desired bond lengths can be obtained by modifying the Morse function and e _σ terms. The π contribution to the L–M–L angle bending operates in the correct sense but is small and can also be accommodated by altering the magnitude of e _σ . For asymmetric π interactions (e _πx ≠ e _πy ) there is no effect on the M–L torsional potential for low-spin d6, high-spin d8 and d9 configurations where the π-symmetry d orbitals are completely filled. Hence, only the σ-bonding contributions to the CLFSE are retained. This approach still gives good agreement with experimental structures, even for formally π-bonding ligands, with average root-mean-square errors in M–L lengths and L–M–L angles of about 0.02 Å and 3° for Co ^III , Ni ^II and four co-ordinate Cu ^II , excluding [Cu(bipy) ₂ ] ²⁺ (bipy = 2,2′-bipyridyl), and about 0.05 Å and 4° respectively for six-co-ordinate Cu ^II , excluding [Cu(terpy) ₂ ] ²⁺ (terpy = 2,2′∶6′,2″ -terpyridyl). The subtle interplay between the axial Ni–Cl and equatorial Ni–N distances in trans-[NiN ₄ Cl ₂ ] macrocyclic species is reproduced for the first time by an MM-based approach. However, the model appears to give relatively poor agreement for [Cu(bipy) ₂ (NH ₃ )] ²⁺ , [Cu(terpy) ₂ ] ²⁺ and [Cu(bipy) ₂ ] ²⁺ . For the five-co-ordinate complex this is due to the intrinsic plasticity of five-co-ordinate copper(II) species. The energy difference between the limiting trigonal-bipyramidal and square-pyramidal geometries is only a few kcal mol ^-1 . For [Cu(terpy) ₂ ] ²⁺ the limiting geometries of tetragonally elongated and compressed octahedra are also within a few kcal mol ^-1 although the present set of parameters overestimates the ligand contribution and predicts a compressed geometry. The calculated structure of [Cu(bipy) ₂ ] ²⁺ is too flat but for four-co-ordinate species it is shown, using [CuCl ₄ ] ^2- as an example, that there are several ways to induce a tetrahedral distortion. The most satisfactory method is to include charges on Cu and the ligand donors whereupon the geometries of [CuCl ₄ ] ^2- and [Cu(bipy) ₂ ] ²⁺ distort to the required flattened tetrahedral structures.