Natural lightning is both frequent and variable and thus a good subject for statistical studies. A typical negative cloud-to-ground (CG) flash consists of multiple individual return strokes. The spatial and temporal distributions of various lightning events throughout the discharge provide a surrogate look inside the CG flash and offer insight into the underlying physical processes. In this study, we combine 8years of National Lightning Detection NetworkTM (NLDN) and North Alabama Lightning Mapping Array (NALMA) data to compute the spatial and temporal distributions of (i) subsequent NLDN-reported return strokes and (ii) LMA-reported sources around NLDN-reported CG strokes. Subsequent strokes are separated into those with the same contact point as the first stroke and those flowing along new lightning channels. Statistically, the distribution of strokes along new channels evolves deterministically, with ∼200 km/s propagation speed from the original channel, comparable to the speed of a stepped leader. This suggests that the −CG subsequent strokes forming new channels may be directly linked to the initial one by a propagating leader inside the cloud. We present LMA case studies and a multiyear analysis of NLDN-LMA data that support this behavior. Our results are supported by ground-truth measurements and video recordings from previous field studies.