Monte Carlo methods apply various physical models, either condensed history (CH) or track structure (TS), to simulate the passage of radiation through matter. Both CH and TS models continue to be applied to radiation dosimetry investigations on a micro and nano scale. However, as there has been no systematic comparison of the use of these models for such applications there can be no quantification of the uncertainty that is being introduced by the choice of physics model. A comparison of CH and TS models available in Geant4, along with a quantification of the differences in calculated quantities on a micro and nano scale, has been undertaken in this study. A sphere of liquid water was simulated, with an incident beam of monoenergetic electrons with kinetic energy between 50 eV and 10 keV. The energy deposition (typical of microdosimetry) and number of ionisations (typical of nanodosimetry), per incident particle, were recorded in a water sphere with diameter varying between 1 nm and 1 m. The simulations were repeated using the following physics packages: Livermore (CH), Penelope (CH) and Geant4-DNA (TS). Results indicated that substantial differences were present between calculated physical quantities, depending on the physics model, target diameter and ratio of the target diameter and mean track length of the incident electron. In the case of the smallest targets, the calculated energy deposition was higher when using the CH models, while the number of ionisations was typically underestimated. In larger targets the energy deposition was in good agreement for all physics models, however the number of ionisations was significantly underestimated by the CH approach, in some cases by almost two orders of magnitude. Regarding CH models, the parameter that had the greatest impact on the results was found to be the threshold of production of secondary particles; when this was minimised the CH and TS results showed the best agreement.