Maths is the physics of bottlecaps. In my experience 2+2=4 is often valid within the limits of observational error, but I am open to the possibility of experimental refutation.