Data retrieved from the Cambridge Structural Database for crystal structures containing metal complexes have been analyzed to evaluate the effects of crystal environment on molecular geometry. Parameters for the same bonds in different crystal sites were compared and variances estimated for these parameters. The primary method used was to compare geometric data for crystal structures for which more than one molecule of complex is present per crystallographic asymmetric unit. Other methods used were comparison of geometry for molecules whose symmetry or pseudosymmetry is higher than the crystal site symmetry, combination of these two methods, and study of compounds for which more than one polymorph has been structurally characterized. These studies allow quantification of the oft-cited "crystal packing" effects on molecular structure. Typically the metal-metal and metal-ligand bond lengths show standard deviations (sigma) on the order of 0.01-0.02 Angstrom apparently due to the effects of crystal environment. The corresponding values for valence angles at the metal atoms typically lie in the range 1-2 degrees although some are over 3 degrees, and for torsion angles the values are still higher (typically 5-10 degrees for cyclic bonds and much greater than 40 degrees for P-Ph bonds in MPPh(3) complexes). The variances of molecular parameters are much larger than the variances estimated from standard crystallographic least-squares procedures and appear to result in large part from crystal packing effects. The implications of these results for the interpretation of crystallographic data are considered. Among consequences discussed are the nature of the structure determined by crystallographic methods, the validity of modeling procedures which mimic crystallographically determined structures, and the relationship between parameters determined by crystallographic and other structural methods.