The lower bound formulae of Temple, Weinstein and Stevenson (1973) are known to give rather poor estimates of the energy when the wavefunction has been calculated by minimization of the Rayleigh upper bound. Recent studies for atoms and molecules have shown that gaussian wavefunctions give particularly poor lower bounds, and this has been explained in terms of the incorrect behaviour of such wavefunctions at and near a nucleus. The authors examine the accuracies of the lower bounds for a series of cusped-gaussian wavefunctions, and discuss the significance of the cusp condition at a nucleus. Numerical evidence for the proposition that -(H2)1/2 may in practice be expected to be a lower bound for a wavefunction calculated by minimization of the upper bound is presented.