We investigate, in the viewpoint of mathematical stability, the validity of the time-strain separability hypothesis employed in polymer viscoelasticity on the basis of experimental results. There have been suggested two distinct stability criteria such as Hadamard related to quick response and dissipative stability conditions, and in the limit of high deformation rate we have proved that separable constitutive equations are either Hadamard or dissipative unstable. The fact that the separability is not valid in the short time region in stress relaxation experiments exactly coincides with the results of our analysis. Therefore, since the application of the separability hypothesis incurs thermodynamic inconsistency as well as mathematical instability, such application should be avoided in the formulation of constitutive equations. In addition, careful attention should be paid to the limit of its validity even in experiments. It is also proved that there is neither theoretical nor physical validity of using the damping function.