In this paper we briefly describe the influence of astrometric binaries on derived proper motions and we draw some conclusions with respect to the reference systems. For more details the reader is referred to Wielen (1997) and Wielen et al. (1997).

Increased measurement accuracy determines not only the derived parameters with much more precision, but it may even allow one to describe new phenomena, which were of no importance before. This happened in the case of the highly precise Hipparcos measurements (ESA, 1997) which were able to determine stellar proper motions within less than four years with an accuracy comparable with the best proper motions available from ground-based observations as given in the FK5 (Fricke et al. 1988). The basic difference between the Hipparcos and the FK5 proper motions showed up within our task of validating the Hipparcos proper motions during the reduction process. Our main result was that the dispersion of the residual differences between both proper motions was about three times as large compared with the error estimates expected from the data given in the catalogues. Since it was rather unlikely that one of the two catalogues had so strongly underestimated its errors, we looked for other reasons to explain the discrepancy. For many of the stars with the largest differences between the Hipparcos and FK5 proper motions, information on their duplicity was found in the literature. We thus suspected that the influence of astrometric binaries to be responsible for the discrepancy.