In crystal structures in which one equivalent position is randomly occupied by different atomic species of different sizes the adjoining atoms must exhibit positional disorder. A special case is the statistical presence or absence of hydrogen atoms as observed in NH4H2PO4 (ADP) and KH2PO4 (KDP), where the bond lengths in the phosphate group will vary locally depending on whether it is a P-O or a P-O-H [P-O(h)] bond. The ideal shape of an individual PO2(OH)2 group is predicted on the basis of the extended electrostatic valence rule and the O-P-O on P-O dependence in orthophosphate groups: P-O 1.501 A, P-O(h) 1.573 A, O(h)-P-O(h) 105"8 °, O(h)-P-O 109-2 ° and O-P-O 113.9 ~. This calculated PO2(OH)2 group is fitted into unit cells of the same dimensions but lower symmetry than those of the paraelectric 1742d structures of ADP and KDP. These hypothetical local environments in C~, 9 (/-centered) and P21212~ are compatible with the observed averaged structure deduced from X-ray diffraction data for ADP. The hydrogen-atom positions as calculated for the least electrostatic energy in the local environments agree with their positions as determined from neutron-diffraction data. The local environments display interatomic distances which differ from the distances in the averaged structure individually as well as in their average values. Therefore, the bond lengths derived from the refinement of averaged disordered structures should be viewed with extreme caution. The distance between the averaged atomic positions can be different from the average of the individual distances. The situation resembles the case of the effects which thermal motion has on the estimation of bond distances. In these instances a valid correction is only possible if we have information on the nature of the thermal motion or on the type of disorder in the crystal structure. Such information is generally not available.