It is well-known that convection electric fields have an important effect on the ionosphere-thermosphere system at high latitudes and that a quantitative understanding of their effect requires a knowledge of the plasma convection pattern. Consequently, convection electric fields have been measured by a variety of techniques, including satellite, rocket, and balloon-borne probes, optical tracking of ionized barium clouds, incoherent scatter observations of drifting F region plasma, and coherent scatter observations of drifting E region irregularities. Since all of these measurement techniques provide information an only a limited spatial region at any time, the construction of the overall convection pattern requires a synthesis of data obtained at a variety of places and times. Some of the problems associated with this procedure are examined in this study using incoherent scatter radar data as an example. In this case the main problem stems from the great variability of the convection electric field. Even for very similar magnetic and IMF conditions, the synthesis of convection data obtained on different days produces a convection pattern which has a fairly large uncertainty. For the case considered, the uncertainty is sufficiently large so that the characteristic differences between the H. Volland [1978> and R. A. Heelis et al. [1982> convection models cannot be clearly identified. In comparison to this uncertainty, the uncertainties associated with assuming how corotation is driven and with transforming velocities between geographic and geomagnetic coordinate systems are negligible. The selection of the ''correct'' magnetic coordinate system is also less important in view of the convection variability.