In the Stern-Gerlach experiment, we have an inhomogeneous magnetic field aligned with the axis labeled z. The two distinct deflections are caused by the force from two distinct magnetic moments [itex]\mu = \gamma S[/itex]. I understand that the spin larmor precesses around the field and that the expectation value for [itex]S_z[/itex] remains constant, but when we collapse the spin by placing a screen to detect to detect the particles, how are we certain that after collapsing the spin state (I think the spin is collapsed before going through the field?) the spin vector cannot change spatially, therefore changing the z component of spin. I suspect my misunderstanding is due to thinking about the "directional" nature of spin the wrong way.