We show that the stress-optic rule is violated in a theoretical model of polymer melts that uses a harmonic potential to confine a polymer chain in a tube-like domain of surrounding entangled polymers. This is the model first proposed by Doi and Edwards (DE) [J. Chem. Soc. Farad. Trans. II 1978, 74, 1802] and later extended by Ianniruberto and Marrucci (IM) [J. Non-Newton. Fluid Mech. 1998, 79, 225]. We first provide a rigorous Green function for such a chain, which is a correction to that derived by DE. The modification guarantees chain continuity. Second, we show that the stress-optic rule is violated in the IM model if the modified Green function is used and an unnecessary assumption employed by IM is avoided. Our result contradicts that of IM. The violation is due to the presence of the virtual springs to confine the chain in the tube rather than the anisotropy of the confinement potential. On the other hand, DE used the virtual-spring model only for estimation of the monomer density along the primitive path where we find just a small correction. As they did not use it for rheological calculations, the stress-optic rule appears to be safe for the Doi-Edwards model. The result might be useful for extracting tube potentials from atomistic simulations. (c) 2014 Wiley Periodicals, Inc. J. Polym. Sci. Part B: Polym. Phys. 2014, 52, 460-469