Data assimilation is a vital technique in modern meteorology, allowing information from a wide range of observations to be combined to produce synoptic analyses of the atmospheric circulation. Data assimilation has a number of advantages over conventional analysis methods that stem from the use of a numerical model of the atmosphere as an integral part of the technique. The model provides a framework which allows information from observations to be combined with our understanding of the atmosphere, as embodied in the model, so that a consistent, evolving, three-dimensional picture of the atmospheric circulation may be constructed. The aim of this paper is to compare results from two different stratospheric data assimilation systems (from Goddard Space Flight Center and the U.K. Meteorological Office) to assess the reliability of analyses derived by using assimilation techniques. In particular, this paper concentrates on assessing the quality of the analyses for long-term transport studies. First, the zonal mean circulations are compared, with emphasis on the consistency of the residual circulation. Second, horizontal transport processes are compared, with particular emphasis on the representation of features in the neighborhood of the wintertime polar vortex. It is shown that the two analyses give a broadly consistent picture of the atmospheric circulation. Where differences are identified, they can often be attributed to particular shortcomings in one of the assimilation systems. The results presented here stress the usefulness of data assimilation for stratospheric studies. ¿ 1997 American Geophysical Union