Ordered weighted differences between matching parameters in independen
tly determined structures, the experimental deviates, follow a Gaussia
n distribution only if both determinations are effectively subject sol
ely to the influence of random effects and all uncertainties in the pa
rameters are reliably evaluated and propagated. Departures from Gaussi
an are readily detectable by plotting experimental deviates against co
rresponding normal probability deviates; a normal distribution produce
s a linear array of unit slope and zero intercept. Application of this
procedure to the results of an X-ray and a neutron diffraction determ
ination of the structure of Ni(ND3)(4)(NO2)(2) at similar to 11 K [Ive
rsen, Larsen, Figgis, Reynolds & Schultz (1996). Acta Cryst. B52, 923-
931] reveals uncertainties in both experiments to be underestimated by
a factor of 2 for the atomic coordinates and 1.5 for the atomic displ
acement parameters (ADPs). In addition, two ADP deviates depart signif
icantly from the remaining twenty that are normally distributed; the d
ifference between X-ray and neutron values for one N-atom ADP is highl
y significant. Uncorrected parameter problems introduced into further
modeling may lead to false inferences.