Stratification in channel belts is the key to reconstructing formative channel dimensions and palaeoflow conditions; this requires an understanding of the relation between river morphodynamics and set thickness. So far, theories for reconstruction of the original morphology from preserved stratification have not been tested for meandering river channels due to the lack of detailed bathymetry. This paper reports the results of an experiment that reproduced a dynamic meandering gravel-bed river with the objectives to: (i) test the prediction of set thickness as a function of the morphology formed by a meandering river channel; and (ii) explore and explain spatial and temporal set thickness variations in the resulting channel belt. High-resolution measurements of time-dependent surface elevation were used to quantitatively relate the preserved stratification to the meandering river morphology. Mean set thickness agrees well with the theoretical prediction from channel morphology. The mean preserved set thickness was 30% of the mean channel depth. Due to the absence of aggradation during the experiment, this provides a lower limit for the preserved mean set thickness which is also to be expected for aggrading systems, because reworking is some orders of magnitude faster than aggradation. Furthermore, the time required to mature a channel belt and its set thickness distribution was about the same as the time required to develop and propagate bends that fill the channel belt surface. Finally, there was much systematic spatial variation in set thickness related to repetitive point bar growth and chute cut-off. Undisturbed and thick sets occurred close to channel belt margins and more irregular stratification with stacked thinner sets was observed in the centre of the channel belt.