Experiments with a limited-area objective analysis are used to determine the effects of various data sources on surface stress and sea ice drift in North Atlantic high latitudes during a 7-day period of the First GARP Global Experiment (FGGE). The objectively analyzed fields provide the input to a planetary boundary layer model, which provides the surface wind stresses for the forcing of a dynamic-thermodynamic sea ice model. Different combinations of drifting buoy, surface marine, and satellite sounder data are deleted from the FGGE level IIB data input. On individual days the stresses are found to differ by as much as 50--100%, and the ice drift speeds by as much as 10--15 km d-1 when buoy reports are deleted from the observing network. Seven-day mean differences of the ice drift speeds are 3--5 km d-1 in some regions. The deviations from the ''complete data'' results are larger when both buoy/marine and satellite data are deleted than when only one type is deleted. While the high-latitude sensitivities to satellite data are generally smaller than to buoy/marine data in the period considered here, large differences in ocean surface stress occasionally result from subtle effects of satellite data on the static stability of the near-surface region. The results also show how data sensitivities can be exaggerated by the imposition of data rejection criteria in experiments utilizing consistent sets of guess fields and input data. The ice velocities obtained from the various experiments are compared with the available buoy displacements and with changes in the ice edge deduced from the Navy/NOAA weekly ice analyses.