This paper addresses the question whether a four-cell convection pattern in the polar cap ionosphere is required by observations, or whether the data are fully explainable by a (perhaps highly distorted) two-cell convection pattern. We present convection data from Atmosphere Explorer C, which, if only the flow component in the sunward-antisunward direction were measured, could be explained either as one of two possible distorted two-cell patterns or as a full four-cell pattern. However, neither of the distorted two-cell patterns that are consistent with the sunward-antisunward flow component can be made consistent with the dawn-dusk flow component over the entire spacecraft trajectory, without postulating a severe flow kink and extra field-aligned currents sunward of the spacecraft track. In addition, the zero potential point (which in a four-cell model would mark the division between the two reverse convection cells) also exactly corresponded to the location of the reversal of the east-west component in the flow, a feature predicted from the four-cell model but more difficult to explain in a distorted two-cell model. Because the pattern was repeated on two consecutive passes, time variations can probably be ruled out as a cause of the sunward flow. Between the two northern hemisphere dayside passes, a southern hemisphere nightside pass also showed a region of sunward flow in the polar cap. The fact that in this case the sunward flow was not confined to the dayside also favors a four-cell explanation.