The general property of fluid motion which is most striking to me is its shifting character. If we vary only one factor of a well‐defined fluid motion problem, for example, if we vary the opening of the flow from a drinking fountain, we can observe a series of phenomena which differ from each other not only quantitatively, but also qualitatively, that is, in their physical nature and appearance.Let us examine an example somewhat closer to our problem of propulsion of sand and other granular media by fluids: consider the flow of an infinitely broad turbulent stream around a smooth obstacle. The detailed data given here refer to a cylindrical body of infinite length, this being the shape of the obstacle most completely studied to date. We can distinguish five qualitatively different types of flow, each having a fairly well‐defined range of Reynolds' numbers (a, b, c, d, e of Fig. 1). The first type is characterized by steady flow, which is of necessity symmetrical with reference to the flow‐direction. With increasing Reynolds numbers. it varies continuously from a flow which is almost symmetrical with respect to the perpendicular to the main flow‐direction, until a stage is reached where a strong unsymmetry is caused by two oblong vortices in a fixed position in the rear of the obstacle (Fig. 1a). But vortices lose their stability when they are very oblong, decomposing into several vortices, and thus giving place to a double row of vortices. Von Karman showed the instability of a symmetrical disposition of two parallel vortex‐rows, explaining why an “antisymmetric” (alternating) pattern of vortices always is found. These antisymmetric vortex‐rows in their turn, for obvious reasons, cannot exist in fixed positions with respect to the obstacle, but must travel downstream. Thus we have here the familiar phenomenon of vortices separating from the body alternately at the sides and giving a perfectly regular pattern of vortex‐filaments, each parallel to the axis of the cylinder (Fig. 1b). At a certain distance downstream from the obstacle they gradually lose their identity, giving place to an irregular, turbulent flow. As Fage and Falkner showed experimentally, this turbulent wake resulting from the disintegration of the double row of vortices shows a three‐dimensional distribution of vorticity and of velocity‐fluctuations in spite of the fact that the double row of vortices itself is, within the accuracy of measurement, a strictly two‐dimensional phenomenon.