Literature data show that the term "squamous-cell head and neck cancer" includes a wide range of epidermoid cell subgroups, each of them with its own intrinsic radiosensitivity (Do values ranging from 107 to 184 in primary tumors, and 146-263 in recurrences; n values ranging from 1 to 5; and, if we consider linear-quadratic model alpha values from 0.273 to 0.490 and beta values from 0.029-0.045). Different sublethal and potential lethal repair times are also observed (4-6 hours and 12-24 hours, respectively), and structural tissue heterogeneity (hypoxic fraction oscillating 5%-30% of the neoplasm). Most important, different kinetic parameters are demonstrated, with Labelling Index ranging from 4% to 30%, phase-S time from 6 to 19 hours, and potential doubling time from 2 to 20 days. On the basis of Fowler's and Barendsen's mathematical models and knowing the potential doubling time and Labelling Index values (derived from bioptic specimens), as well as alpha/beta ratio for both tumor and normal tissue, we tried to identify the optimal fractionation (standard, accelerated, hyperfractionated) for slow/fast-growth tumors, also evaluating the relative acute and late side effects. Our analysis shows that: 1) tumors with aggressive biological behavior (Labelling Index greater than 15%, aneuploidy, potential doubling time less than 5 days) seem to respond to accelerated fractionation/hyperfractionation without split better than to standard regimens: 2) tumors with slow growth (Labelling Index less than 15%, potential doubling time greater than 5 days, euploidy) seem to respond not only to standard regimens, but also--and mainly--to hyperfractionation.