This sounds almost too obvious, but when you approximate a function locally by a parabola and then solve for its intersection with a line (which occurs in a lot of contexts, e.g. ODE solving), there would seem to really be no interpretation of the square-root except as the root of a 2nd-degree polynomial...