We reexamined the mechanism of the unimolecular rearrangement connecting both nitromethane and methyl nitrite isomers. The CH3NO2 potential energy surface was constructed using different molecular orbital [CCSD(T) and CASSCF] and density functional theory (B3LYP) methods including a few lower lying isomeric intermediates. A particular attention has been paid to the two following questions left open by earlier experimental and theoretical studies: (a) does the interconversion between nitromethane and methyl nitrite by a 1,2-CH3 migration occur via a "loose" or "tight" transition structure (TS)? and (b) is the energy barrier associated with methyl migration actually smaller or larger than the C-N bond dissociation energy? The C-N bond dissociation energy was evaluated with BDE(CH3-NO2) = 60+/-2 kcal/mol in line with available results. In contrast to earlier studies (McKee, M. L. J. Phys. Chem. 1989, 93, 7365, and Saxon, R. P.; Yoshimine, M. Can. J. Chem. 1992, 70, 572) but partly in agreement with a recent G2MP2 study (Hu, W. F.; He, T. J.; Chen, D. M.; Liu, F. C. J. Phys. Chem. A 2002, 106, 7294), our multiconfigurational CASSCF computations demonstrated that the methyl migration involves a "tight" TS whose electronic wave function is dominated by the Hartree-Fock configuration. Calculations are thus internally consistent indicating that the energy of the TS for 1,2-CH3 shift is at least 6 kcal/mol above the CH3 + NO2 asymptote. Thus, a discrepancy with a previous evaluation of experimental findings (Wodtke, A. M.; Hintsa, E. J.; Lee, Y. T. J. Phys. Chem. 1986, 90, 3549), which placed the CH3 + NO2 limit by 5 kcal/mol above the rearrangement TS, appears to persist.