We have used high speed streaking photographic techniques to time-resolve the luminous components of cloud-to-ground lightning flashes. All recordings were made during our participation in the Thunderstorm Research International Program (TRIP), conducted at the Kennedy Space Center, Florida, during the summers of 1977 and 1978, and at the Langmuir Laboratory, near Socorro, New Mexico, during the summer of 1979. Twenty one dart leaders, four dart-stepped leaders and three stepped leaders were recorded, the majority under daylight conditions. The mean two-dimensional propagation speed of the dart leaders, evaluated over channel lengths less than or equal to 0.8 km above ground, is 11¿106 m/s, with a range of 2.9 to 23¿106 m/s. Several of the dart leaders reveal a decrease in propagaton speed as ground is approached. However, four of the dart leaders in two separate flashes show an increase in speed near the ground, an observation not previously reported in the literature. In two multistroke flashes, we examine the variation of dart leader propagation speed along the channel and find very similar behavior for different strokes in the same flash. The speed variations that we observe may be predominantly caused by geometrical variations of the channel. The dart leader propagation speeds reported in this study are compared with the earlier works of Schonland, McEachron, and Kitagawa and Brook. Agreement among the studies is good, with a common range of observed dart leader propagation speed of 2 to 23¿106 m/s. The major discrepency among these studies is the observation, by Schonland, of a distribution of dart leader propagation speeds positively skewed toward the lower limit of reported values. Eleven of the dart leaders are analyzed at upper and lower levels along the visible channel to give 22 dart lengths. They range from 7 to 75 m with a mean of 34 m. For these 22 determinations, we calculate a correlation coefficient of 0.85 between the dart length and the dart leader propagation speed. The correlation of greater dart length with higher propagation speed is consistent with the slower decay of channel luminosity due to the greater initial input of energy to the channel by the faster and, presumably, more energetic dart leader. Four dart-stepped leaders are examined in detail with regard to variation of propagation speed, step length, stepping interval, and luminous intensity during propagation between the cloud base and ground. Significant differences in the tendencies of these parameters are found within these four leaders. For example, one dart-stepped leader recording shows a decreasing propagation speed and an increasing step interval near ground, whereas another shows the opposite behavior. In the best event recorded, several of the individual steps reveal a photographic film density structure, with the lower portion of the step exhibiting a distinct, bright tip that fans out into a symmetrically diffuse image in the upper portion of the step. Our analysis indicates that this spread in the upper portion of the step image is not the result of streaking photography distortion but, rather, represents the luminous structure of the step. We estimate that the step image is recorded in less than 1μs. Consequently, with a measured step length of ~20 m, the luminous pulse must propagate along the step at a speed of at least 2¿107 m/s. The mean propagation speed for three stepped leaders is found to be 1.1¿106 m/s. All three stepped leaders are very faint, and were recorded only in the last 100--200 m above ground. Two stepped leaders and one dart-stepped leader do not propagate completely to ground before initiation of the return stroke. Apparently, these leaders are met by an upward propagating discharge at heights above ground of 20, 30, and 20 m, respectively. Other stepped and dart-stepped leaders cases are indeterminate because an obstacle or the horizon prevent the recording of the leaders near the ground. Connecting discharges are not observed for any of the dart leader events with a resolution of 10 m at 5 km, implying that upward discharges initiated by the approach of dart leader do not occur or are substantially less than a few tens of meters in length. Dart leaders, apparently, propagate completely to ground.