A SIMS procedure has been developed to quantify H, Li and B in schorl-elbaite series tourmalines. It relies on the use of calibration curves determined from various tourmaline standards, mainly characterized by crystal-structure refinement (SREF) and electron-microprobe analysis (EMPA). The SIMS data have allowed us to investigate matrix effects on H, Li and B ionization (with Si and Al assumed, in turn, as the matrix-reference element), and to test the accuracy of our general calibration curves, previously developed for silicates. In particular, the H/Si as well as H/Al ionization behaviour are not indicative of significant residual matrix effects; the agreement between SIMS and reference data is mostly within the reproducibility of analysis (2sigma ), and for all samples is within 10%. The IY(Li/Si) and IY(Li/Al) are affected by matrix effects possibly related to the different (Fe+Mn) and/or (Si+Al) content of the sample. Since the maximum variation in terms of chemical composition in these matrices concerns Fe and Mn concentrations, we used at a first approximation (Fe+Mn) content to investigate the IY(Li) variation among the various samples. A linear regression: IY(Li/Si) (or similarly, IY(Li/Al) vs. (Fe+Mn) (cat %) was obtained and used to improve the accuracy of analysis. All the experimental IY(Li) agree with the expected values within 13% rel. (in one case: 24% rel.), mostly comparable to the uncertainty of Li estimation in these matrices. We can pessimistically put the uncertainty of our regression lines for Li at 15-20% rel. The IY(B/Si) and IY(B/Al) seem to be affected, although to a lower extent than Li, by the same kind of matrix effects, the maximum variation of IY(B) being approximately 10-12%. Accurate B analyses, better than 3% rel., can be obtained by calibrating IY(B) vs. (Fe+Mn) content of each sample. A few examples of this procedure testify the possibility to use this approach to achieve accurate data for H, Li and B by SIMS in schorl-elbaitic tourmaline samples.