Characteristics of stroke occurrence are investigated for lightning to ground observed near Tampa, Florida (latitude=27.4¿N, longitude=82.2¿W) during three days of convective thunderstorm activity in July 1979. Strokes were identified from both wideband (0.2 Hz to 1.5 MHz) electric field records and simultaneous video tape recordings from a network of TV cameras. The average number of strokes per flash was 4.0 for 105 flashes and the average number of channels per flash was 1.6 for 78 multiple-stroke flashes. Interstroke intervals were distributed according to a lognormal rather than a normal distribution with geometric mean of 69 ms and standard deviation of 0.32 decades in the logarithmic distribution for 310 intervals. Interstroke intervals preceding new channels had a geometric mean of 83 ms and standard deviation of 0.33 decades in the logarithmic distribution for 28 intervals. By comparing our results with others in Florida and New Mexico, we could find no obvious relationship between these intervals and channel length. Historgrams of interstroke interval plotted as a function of stroke order did not appear to reveal the systematic increase noted by workers in Japan. This conclusion is corroborated here on the basis of the same statistical tests used by the Japanese workers. The effectiveness of videotape records for determining stroke occurrence in daytime lightning is considered. although 83% of strokes were detected on the videotape record, for ground flashes with at least one stroke visible on TV, there was a significant difference at the 1% confidence level for the t test between the interstroke interval distribution obtained using both wideband electric field and videotape records and that obtained using only videotape records.