“…The main sources of model bias in multipolar density studies are: the choice of exponents appearing in the radial parts of the deformation functions, still`more of an art than a science' (Flensburg et al, 1995); the insuf®cient radial¯exibility in modelling valence-charge density in metals, minerals (Nowack et al, 1991;Brown et al, 1993) and coordination complexes (Iversen, Larsen, Figgis & Reynolds, 1997);² and the limited order of the spherical harmonics used, which do not usually extend past the hexadecapolar level l 4. The latter limitation is not always imposed by the quality of the data, and can be due to the need to preserve an adequately high observations/parameters ratio.…”