In radiological protection, models are used to assess radiation risk by means of extrapolation from high dose and dose rate to low dose and dose rate. In this thesis five main biophysical models of radiation action have been evaluated, appraised and inter-compared. The five models are lethal and potentially lethal (LPL) by Curtis, pairwise lesion interaction (PLI) by Harder, cellular track structure (CTS) by Katz, hit size effectiveness (HSE) by Bond and Varma and track core (TC) by Watt. Each model has been developed based on certain underlying mechanisms or phenomena, to permit interpretation and prediction on the induction of a specified biological endpoint such as cell reproductive death, chromosome aberrations and mutations. Biological systems of interest are, for example, mammalian cells containing deoxyribonucleic acid (DNA). Evidence is mounting that double strand breaks in the DNA are the critical lesions for various biological end points. To proceed with this work the TC model has been chosen. Cancer induction by ionising radiation is the stochastic effect of prime concern in radiological protection. Cancer induction cannot be avoided entirely but its frequency of occurrence may be reduced to acceptable level by lowering the amount of radiation received. The methods of assessment developed by ICRP, in terms of the cancer risk coefficients, are presented in this thesis. In the conventional (legal) system of dosimetry, radiation is quantified by the amount of energy absorbed per unit mass of tissue. Quality factors, superseded by radiation weighting factors, are needed to account for the quality dependence on radiation type. As an alternative, a new dosimetry system is proposed here which is based on the mean free path for primary ionisation along particle tracks and the integral fluence generated by the radiation field, whether directly or indirectly ionising radiation. From the study of cellular data, the mean free path for primary ionisation along particle tracks (lambda) emerges as a parameter which best unifies biological damage data. Radiation effect is found to depend, not on the energy transferred but to depend mainly on the frequency and spatial correlation of interactions. Maximum effect occurs when lambda is equal to lambda0 (2 nanometre, nm). The term 'Absolute Biological Effectiveness' (ABE) is introduced as a parameter which indicates the probability to induce a specified effect, per unit incident fluence. In this endeavour, only direct effects are considered in deriving ABE values for various radiations. However other factors such as indirect effects, inter-track action, repair processes and radiation rate, can be incorporated later if required, in the derivation of ABE. ABE values for photons up to 60Co i.e 1253 keV and neutrons up to 105 keV, have been calculated and presented in this thesis. An attempt has been made to re-express the cancer risk coefficients, derived by ICRP, in the new dosimetry system, in terms of the ABE (Absolute Biological Effectiveness). The hypothesis put forward in this thesis is that the induction of a specified biological-end-point in a biological system due to ionising radiations, is determined not by the amount of energy absorbed per unit mass (dose), but rather by the number of events (ionizations) spatially correlated, along the primary radiation track. Based on this hypothesis, a new unified dosimetry system, independent of radiation type, is proposed. Suggestions are made for possible measuring instruments which have the equivalent response characteristics, namely maximum efficiency of detection for the mean free path Success in devising such types of instrument would ensure the practicability of the new dosimetry system, in operational radiological protection.