The determination of the (Fe, Mg)-cation distribution in important roc
k forming minerals such as olivines and orthopyroxenes from single cry
stal X-ray diffraction data is well established. The findings obtained
from good quality data are, however, subject to the structure model,
refinement strategy, data weighting etc. Moreover, they do not necessa
rily agree with those of comparable Mossbauer studies. In order to ass
ess the reliability of the X-ray results, i.e. precision and accuracy
of the site occupancies and ensuing cation distribution coefficients K
-D, an interlaboratory study was undertaken. Two olivine and two ortho
pyroxene data sets, one experimental and three synthetic, were dispatc
hed to interested applicants for structure refinements according to th
eir own choice. Approximately correct structure parameters were given,
but no information about the chemical composition, i.e. the Fe-conten
t, was communicated. The total of 25 refinements for each data set, ca
rried out with different programs, valence models, and weighting schem
es, yielded some considerable differences between the results, which a
re summarized and discussed in terms of precision and accuracy. The un
certainties of the refined site occupancies, following from the qualit
ies of the fits, are only realistic, when they are dominated by the st
atistical errors in the data. Errors in the independently refined Fe-c
oncentrations on the M1-and M2-sites are not fully transferred to the
thermodynamically relevant cation distribution coefficient. Therefore,
correlations between the site occupancies, clearly recognizable from
the compilations of all respective results, can be combined with an in
dependently obtained chemical constraint (microprobe analysis) in orde
r to yield reliable K-D-values with realistic uncertainties. In summar
y, all results indicate the need for multiple refinement techniques, i
f reliable results on the absolute scale are desired.