In 1997 the Jamuna Bridge was finished and came into use. As this bridge is a shore connection for one of the largest rivers in the world, extensive river training works were designed and built, mainly to avoid outflanking of the bridge by the river. It was especially the braiding character of the river with its ever-changing pattern, which made it difficult to predict the morphological behaviour. The river training works incorporated two guide-bunds, which lead the river through the contraction at the location of the bridge. This made it possible to cross the flood plain by constructing earthen abutments, leaving a bridge span of 5km instead of 15km. The narrowing of the river will lead to morphological changes. Deepening of the riverbed by means of local scour can be expected. To protect the toe of the revetment against this scour, a more or less traditional solution has been applied: the falling apron. A falling apron is a stack of granular material at the toe of a revetment, which will launch onto the slope of the scour hole when the revetment is undermined. The established slope will then be protected by a layer of rock, which will retain the bed material and prevent the formation of a slope that is too steep. To understand more about the setting process and to determine the resulting slope, a physical model was built and tested in a flume. Three different rock sizes were examined and two different initial configurations. It was found that the slope angle after setting was 1:2 for all experiments. Surprisingly, only a small amount of apron rock was actually used. This was because just a single layer of material covered the scour slope. This single layer does not retain the material like a properly designed filter, but it does slow down the erosion of sand from under the apron. Two parts of the falling apron were discerned: the waiting part consisting of the stack of unused material and the 'launched part', which is the rock layer covering the slope. This single layer has important consequences for the functioning of the apron. The grading used at the Jamuna Bridge was extra wide. The idea was that when a thick layer was launched on the slope, the fines would wash out of the top layer. A filter structure would then develop with large grains in the upper layer and fines (and large grains) in the lower layer. This theoretical case was not found in the model. The resulting slope appeared to have settled evenly and over the complete scour depth. No unprotected sand slope was observed during the experiments. The rock layer follows the bed level and extra rock is added from the waiting part of the apron when the layer slides down. This mechanism of scour at the toe and rock launching at the top implies that sand transport through the layer also occurs. The consecutive phases in the setting process showed that at different times the apron slopes lie parallel and retreat as a whole with the edge of the waiting apron. The survey data, retrieved since the Jamuna Bridge was built, were used to compare the model results with a prototype. These profiles also showed an apron slope of 1:2. However the thickness of the layer cannot be determined from the data. It can be seen that the apron slope follows the edge of the waiting part as it does in the model.