During a campaign in January 1988 we observed auroral arcs with the EISCAT UHF radar system and a low light level TV camera. The aim was to compare apparent motions of the arc and its small-scale structures with plasma motions in the adjacent F-region. It was found (1) that there was a relative motion between arc and plasma of the order of 100-200 m/s southward and (2), in confirmation of earlier findings, that folds and curls, i.e. auroral rays seen edge-on, with wavelengths ranging from 2 to 20 km, moved along the arc with average velocities of 10 to 20 km/s. At the same time, the tangential plasma velocity at 300 km height varied between - 0.3 and 0.8 km/s westward. Most importantly, for all luminous bands exhibiting fast motions of small-scale structures, the direction of this motion was westward on the equatorward edge, and on most occasions, an eastward directed counterflow could be observed on the poleward edge. The existence of such a counterflow appears to be a fundamental property and may even be used for a physical definition of an auroral arc. We support the interpretation of Davis (1978) that auroral ray, fold, or curl motion is a visible expression of a physical motion of small-scale structures in the local plasma environment of the auroral acceleration region at about 1 R E altitude. The attendant electric fields are shorted out by parallel potential drops. The counterflow pattern is consistent with the existence of a U-shaped potential at that altitude. The observed asymmetry in brightness, definition, and speed of the counterflowing folds or curls is attributed to their location at the leading or trailing edges of the arc propagating relative to the plasma frame.