Previously, from dielectric relaxation measurements obtained for various temperatures T and pressures P on polypropylene glycol, 1,4-polyisoprene, and poly(oxybutylene) [Roland, C. M.; Casalini, R.; Paluch, M. J. Polym. Sci. Polym. Phys. Ed. 2004, 42, 4313. R. Casalini and C. M. Roland, Macromolecules, 2005, 38, 17791, both the primary a-relaxation time τ(a). and the normal mode relaxation time τ(n). were shown to yield master curves when plotted versus the quantity T-V-1-(γ). Moreover, the value of γ is the same for the two processes (here V is the specific volume and y is a material-specific constant). Such a result appears to be consistent with an assumption underlying models for polymer viscoelasticity that the friction coefficient governing motions over large length scales can be identified with the local segmental friction coefficient. However, notwithstanding the superpositioning obtained using the same value of γ, τ(a) and τ(n) differ in their dependences on either the product variable T-V-1(-Y) , V at constant T, or T at constant P. In each case, the difference is more pronounced at longer τ(a). Such behavior is inconsistent with the Rouse and tube models. However, these differences in the respective dependences of τ(a) and τ(n) can be accounted for quantitatively by the coupling model. The framework of the solution of the problem supports the proposal that the temperature and volume dependences of molecular mobility, which trigger the glass transition, do not originate from the primary a relaxation. Instead, they have their origin in the primitive relaxation of the coupling model.