Observations of solar magnetic and velocity fields can be used to derive the course of events involved in the solar cycle. These differ in three important respects from those of conventional dynamo theories: (i) Polar field reversal. Following the outbreak of a new cycle, magnetic flux released by sunspots diffuses initially by Leighton's random-walk process, but this is soon dominated by the observed poleward flow of about 20 m s - 1 which carries flux to polar regions in about 12 months. Since follower spots lie about 2� higher in latitude than leaders, follower flux arrives in polar regions some two weeks ahead of leader flux, providing a net inflow of follower polarity there until sunspot maximum, reversing the polar field from the previous sunspot cycle and building it up to a maximum. After sunspot maximum, the flux arriving in polar regions is predominantly of follower polarity until or unless spots occur at latitudes so low that flux can diffuse towards and across the equator, predominantly from the lower latitude leader; the effect is doubled by a complementary migration from the opposite hemisphere. This prevents the change in polar flux over the cycle from dropping to zero, and leaves the polarity there reversed at the end of the cycle. (ii) The sunspot cycle. A slow, deeper counterflow, essential for continuity, carries flux strands down in the polar zones and then equatorwards. The concentration of strands is increased continually by differential rotation, and they are dragged continually into contact. Reconnection occurs rapidly except between tubes that are inclined at very small angles. This results in the formation of ropes of flux strands twisted very gently. At some stage they are large enough to float, forming sunspots. The mean sunspot latitude decreases continuously as the flux is carried equatorwards, dying out as the flux ropes become exhausted. The whole process repeats, once again reversing the polar and spot group magnetic fields. Hale's polarity laws follow immediately, and Sporer's law requires only minor adjustments to the predicted velocity of the deep equatorward counterflow. The estimated velocity of this flow is compatible with the observed sunspot and magnetic cycles of 11 and 22 years. (iii) The torsional oscillation. Shear by differential rotation increases the concentration of flux strands; the reaction to strongly sheared flux strands is a tendency to reduce differential rotation. This results in cyclic variations of differential rotation, the phase with respect to sunspot formation being in good agreement with the torsional oscillation observations of Howard and LaBonte (1981) at all latitudes up to 50-55�.