Fig 3. Failure of a single-channel model to account for the spreading rate in Goldsmith’s data.

A: A plausible single channel model gives a reasonable fit at zero time. B: The same model gives a very poor fit at 30 mins because it cannot match the spreading rate. The model graphs have been adjusted so the peaks of model and data are the same height, to facilitate comparison of their widths. The data for panels A and B were obtained by measuring histograms in Fig 1A and 1D, respectively, in [13].