Introduction of heavy particles (hadrons) into radiation therapy aims at improving the physical selectivity of the irradiation (e.g. proton beams), or the radiobiological differential effect (e.g. fast neutrons), or both (e.g. heavy-ion beams). Each of these new therapy modalities requires several types of information before prescribing safely the doses to patients, as well as for recording and reporting the treatments: (i) absorbed dose measured in a homogeneous phantom in reference conditions; (ii) dose distribution computed at the level of the target volume(s) and the normal tissues at risk; (iii) radiation quality from which a RBE evaluation could be predicted and (iv) RBE measured on biological systems or derived from clinical observation. In hadron therapy, the RBE of the different beams raises specific problems. For fast neutrons, the RBE varies within wide limits (about 2 to 5) depending on the neutron energy spectrum, dose, and biological system. For protons, the RBE values range between smaller limits (about 1.0 to 1.2). A clinical benefit can thus not be expected from RBE differences. However, the proton RBE problem cannot be ignored since dose differences of about 5% can be detected clinically in some cases. The situation is most complex with heavy ions since RBE variations are at least as large as for fast neutrons, as a function of particle type and energy, dose and biological system. In addition, RBE varies with depth. Radiation quality thus has to be taken into account when prescribing and reporting a treatment. This can be done in different ways: (a) description of the method of beam production; (b) computed LET spectra and/or measured microdosimetric spectra at the points clinically relevant; (c) RBE determination. The most relevant RBE data are those obtained for late tolerance of normal tissues at 2 Gy per fraction ("reference RBE"). The "clinical RBE" selected by the radiation oncologist when prescribing the treatment will be close to the reference RBE, but other factors (such as heterogeneity in dose distribution) may influence the selection of the clinical RBE. Combination of microdosimetric data and experimental RBE values improves the confidence in both sets of data.