Abstract
A simple correction accounting for the effect of truncating the atomic orbital basis on the calculation of N.M.R. shielding is proposed. This correction is obtained from the expression for the magnetic shielding as a sum of orbital pair contributions: a spurious term that vanishes only if a complete basis is used appears in this expression. It is demonstrated both theoretically and numerically (using the example of PH3) that removing this term from the results obtained with a truncated basis improves significantly the agreement with the experimental values. As a by-product the gauge dependence of the numerical results is much reduced. The additional computing time is negligible.