A critique of the instantaneous normal mode (INM) theory of diffusion by Gezelter, Rabani, and Berne (GRB) [J. Chem. Phys. 107, 4618 (1997)] is analyzed. GRB assert that imaginary-frequency INM are corrupted with modes unrelated to barrier crossing, that proposals for removing such nondiffusive (ND) modes are inadequate, and thus that INM cannot be used to predict the self-diffusion constant, D. In rebuttal it is argued that Lennard-Jones, the system studied by GRB, is anamolously rich in ND modes. INM in molecular liquids are shown to behave as excellent indicators of barrier crossing. Even in LJ ND-INM, while plentiful, do not dominate D except in supercooled liquids near Tg and in the crystal; hence the many successes of the theories already reported in LJ. Agreement of simulated Im-ω densities of states with calculations modeling the INM as excitations in the soft potential model also indicate that the INM reflect the potential energy landscape in liquids.