Radio frequency observations of cloud-to-ground lightning in Florida have been analyzed to document a number of features of the lightning. The observations have been made using an interferometer system which determined the direction to the lightning radiation as a function of time during close lightning discharges. Observations are presented for about 50 radiating events during five multiple-stroke, normal-polarity flashes to ground. The results confirm and extend results of a similar study of New Mexico lightning by Rhodes et al. (1994). Dart leaders, in-cloud “K” events, and attempted leaders are found to be the same phenomena, namely, a negative-polarity streamer that propagates horizontally and/or vertically to ground at estimated speeds of about 107 ms−1 down to 106 ms−1. The step or impulsive fast electric field change produced by some K events is found to be caused by positive breakdown back at the starting point of the negative streamer, after the streamer has progressed over some or all of its distance. This breakdown appears to initiate a “forward” stroke along the streamer channel which can renew the breakdown at the front end of the channel. K breakdown which happens to occur during a continuing current discharge to ground produces an M event (channel brightening) when it connects with the conducting channel to ground. The step fast field change of the M events is found to be produced at the time of connection inside the cloud; no fast field change is detected when this breakdown reaches ground that would be associated with a return stroke. M events can also be initiated by fast (107 ms−1) positive streamers which often propagate away from the leader source region within a few milliseconds of the arrival of a return stroke back in the source region. These streamers often generate an even faster negative recoil event back along their extent which propagates down the channel to ground, as the M event. In all cases the current increase or channel brightening produced by K and M events is predicted to occur in a forward direction, that is, toward or down the channel to ground, in the same direction as the initiating negative streamer. Positive breakdown is detected only immediately following return strokes or during some K streamers. In both cases the electric stresses are expected to be very large, suggesting that positive breakdown which produces a conducting channel is difficult to initiate. Results for one flash suggest that negative charge was left at low altitude by the initial stroke to ground; this forced the subsequent leaders to develop a new path to ground, first as series of attempted leaders and then as a dart-stepped leader.