Order α 2 contribution to the orthopositronium decay rate due to one-photon virtual annihilation is found to be δannΓ (2) = α π 2 π 2 ln α − 0.8622(9) ΓLO.PACS numbers: 36.10.Dr, 06.20.Jr, 12.20.Ds, 31.30.Jv Positronium, the bound state of an electron and positron, is an excellent laboratory to test our understanding of Quantum Electrodynamics of bound states. Although in the majority of cases the agreement between theory and experiment is very good, the case of orthopositronium (o-Ps) decay into three photons is outstanding, since the theoretical predictions differ by about 6 standard deviations from the most accurate experimental result [1] (see, however, an alternative result in [2]). Provided that the experiment [1] is correct, the theory can only be rescued if the second order correction to the o-Ps lifetime turns out to be ∼ 250(α/π) 2 Γ LO . It is however difficult to imagine how such a large number could appear in the perturbative calculations, even if the bound state is involved. This point of view is supported by a recent complete calculation of the O(α 2 ) correction to the parapositronium (p-Ps) decay rate into two photons [3]. It has shown that the "natural scale" of the gauge-invariant contributions is [several units]×(α/π) 2 Γ LO . For this reason, it was conjectured in [3] that the O(α 2 ) correction to the orthopositronium decay rate o-Ps → 3γ most likely is of the same order of magnitude.At first sight, the result of Ref.[4],for the gauge-invariant contribution to the o-Ps→ 3γ decay rate induced by the single-photon virtual annihilation, does not provide much support for this conjec- * e-mail: g adkins@acad.fandm.edu † e-mail: melnikov@particle.physik.uni-karlsruhe.de ‡ e-mail: yelkhovsky@inp.nsk.su ture. In fact, the value of the non-logarithmic constant in Eq. (1) is larger by approximately one order of magnitude than the values of coefficients in gauge-invariant contributions to p-Ps→ 2γ decay rate. In this note we would like to point out that the result for the second order correction to virtual annihilation contribution given in Eq. (1) is incomplete, in that closely related contributions should be included as well. It turns out that if the missing pieces are added to Eq.(1), then the complete result for δ ann Γ (2) decreases and is in accord with the expectations advocated in Ref. [3].We recall, that in bound state calculations there are two different types of contributions. The hard corrections arise as contributions of virtual photons with momenta k ∼ m. These contributions renormalize local operators in the non-relativistic Hamiltonian. For this reason they can be computed without any reference to the bound state.On the contrary, the soft scale contributions come from a typical momenta scale k ∼ mα in virtual loops, and for this reason are sensitive to the bound state dynamics. For δ ann Γ (2) , it is easy to see that the soft scale contribution reads:whereis the reduced Green function of the Schrödinger equation in the Coulomb field.Let us write the expansion of the Green function ...