Auroral spirals are the largest vortex structures that appear in auroral arcs. Their diameters vary from tens to hundreds of kilometers. We have collected an all-sky camera data set of 216 events including 189 northern and 27 southern hemispheric spirals. For the first time in the literature, this study shows spiral structures from both hemispheres with similar occurrence distributions for their scale sizes and a sense of rotation that is always counterclockwise when viewed along the geomagnetic field. In a few cases the time resolution was adequate to follow the winding process and the relative brightness of the spirals. These events show that the winding motion of the auroral structure is opposite to the anticipated plasma flow in the same region, and that the winding process is controlled by the intensity of the auroral precipitation. We show that these findings are consistent with the model of Hallinan [1976>. He suggested that spirals are related to periodic undulations developing in an upward field-aligned current (FAC) sheet. In this model, the winding of the auroral structure manifests the magnetic field line twisting in a direction different from the ionospheric plasma flow. We discuss Hallinan's theory in more detail: First, we derive the complete solutions of the magnetic field perturbation. Second, we compute the structure of the FAC sheet and show that it reproduces the shape and sense of rotation of auroral spirals. Third, we consider the factors that affect the amount of spiral winding. In particular, we show that the spiral winding is controlled by the strength of the FAC perturbation in the arc and thereby is related to auroral precipitation and arc brightness observed. We also find the relationship between the finite field-aligned distance required for spirals to fully evolve and the amplitude of the FAC perturbation and discuss how this relationship could be used to identify the source region for the FAC perturbation. However, the assumption of a constant background field in the model hampers detailed comparisons with observed FACs and spatial scale sizes. Further development of the model is needed before we can conclude, for example, if at the altitude of the acceleration region, the conditions would be favorable for the spiral winding mechanism. ¿ 2001 American Geophysical Union