A radiation action model based on nanodosimetry is presented. It is motivated by the finding that the biological effects of various types of ionizing radiation lack a consistent relation with absorbed dose. It is postulated that the common fundamental cause of these effects is the production of elementary sublesions (DSB), which are created at a rate that is proportional to the probability to produce more than two ionisations within a volume of 10 base pairs of the DNA. The concepts of nanodosimetry allow for a quantitative characterization of this process in terms of the cumulative probability F 2 . The induced sublesions can interact in two ways to produce lethal damage. First, if two or more sublesions accumulate in a locally limited spherical volume of 3–10 nm in diameter, clustered DNA damage is produced. Second, consequent interactions or rearrangements of some of the initial damage over larger distances (~ µm) can produce additional lethal damage. From the comparison of theoretical predictions deduced from this concept with experimental data on relative biological effectiveness, a cluster volume with a diameter of 7.5 nm could be determined. It is shown that, for electrons, the predictions agree well with experimental data over a wide energy range. The only free parameter needed to model cell survival is the intersection cross-section which includes all relevant cell-specific factors. Using ultra-soft X-rays it could be shown that the energy dependence of cell survival is directly governed by the nanodosimetric characteristics of the radiation track structure. The cell survival model derived in this work exhibits exponential cell survival at a high dose and a finite gradient of cell survival at vanishing dose, as well as the dependence on dose-rate.