Liquid phase vibrational energy relaxation (VER) times T1 typically depend critically on the relaxing mode's high frequency friction or wing function. The wing function may, in principle, be found from the mode's normalized force autocorrelation function (faf) C(t), since it is proportional to lim formula see text C(t)dt. However, the full form of C(t) is never available. Thus, the wing function is typically estimated from a model faf CM(t) which duplicates the known part of C(t) and which (hopefully) approximates its unknown part with enough realism to yield meaningful formula see text behavior. Unfortunately, apparently realistic CM(t)'s can predict unphysical wing functions, and T1's in error by tens of orders of magnitude. Thus, a condition is needed to discriminate between CM(t)'s which yield meaningful and unphysical forms for the high frequency friction. This condition is shown to be that model faf's CM(t) yield physical wing functions if and only if these functions derive from the short time "heads" of the faf's. This test is applied to the model faf's Cga(t) triple bond exp[-1/2(t/tau)2] and Cse(t) triple bond sech(t/tau). These faf's cannot both be physical, since they yield incompatible Gaussian and exponential wing functions. The test accepts Cga(t) as physical. It, however, rejects Cse(t), since its "tail" lim formula see text Cse(t) = 2 exp(-t/tau) (because of its long range) dominates the wing function.