We test the two most commonly used lightning return stroke models. Bruce-Golde and transmission line, against subsequent stroke electric and magnetic field wave forms measured simultaneously at near and distant stations and show that these models are inadequate to describe the experimental data. We then propose a new return stroke model that is physically plausible and that yields good approximations to the measured two-station fields. Using the new model, we derive return stroke charge and current statistics for about 100 subsequent strokes.