Rivers meander, and the conventional explanation is that an obstruction intervened. A hard outcrop, a fallen tree, a bend inherited from some antecedent course: the channel is deflected, and once deflected it remains deflected because the outside of a bend erodes and the inside deposits. The explanation is not erroneous about the mechanism. It is wrong about what needs explaining.
Meanders form in materials containing no obstacles whatsoever. They form in laboratory flumes of uniform sand laid dead flat, in glacial meltwater traversing the surface of ice, and in the ocean, where the Gulf Stream meanders through water. In every one of those instances the medium is homogeneous and the pattern appears regardless, which signifies that the pattern is not a response to anything in particular. It is what a flow of this description does when nothing prevents it.
Why it does so is a question concerning energy. Water descending a slope has energy to dispose of, and a channel that carried it straight would dispose of it entirely at the outlet, in a manner the bed could not withstand. A meandering channel is longer, so its gradient is gentler, and consequently the energy is dissipated gradually along the whole length instead of being concentrated. The curves themselves are not arbitrary: measure sufficiently many of them and the radius of a bend emerges at two or three times the channel width, in rivers of every dimension, from a metre across to a kilometre. A pattern that scales that consistently across four orders of magnitude is not an accumulation of accidents.
The engineering consequence is extensively documented and was learned expensively. Rivers straightened for navigation or for flood control acquire a steeper gradient over the shortened reach, and the current accelerates, and the bed scours. Downstream of the works the sediment that was removed upstream is deposited, and the channel there shallows and floods more readily than previously. Left alone, a straightened reach commences re-meandering within a few seasons, and the customary response has been to armour the banks -- which fixes the plan but not the energy, and moves the problem to wherever the armouring ends.
None of this renders obstacles irrelevant. An outcrop determines where a particular bend sits, and a fallen tree can initiate one in a given season. What obstacles do not do is explain why a river bends at all, and the conventional account has been answering the second question with material that only bears on the first. The distinction is worth preserving because the two questions have different answers and different applications. Where a bend sits can be predicted from a map of the underlying rock. Whether there will be bends at all can be predicted from the gradient and the material, and from nothing else.
