The increasing complexity and heterogeneity of modern light water reactor (LWR) fuel assemblies impose new challenges to current reactor physics codes in terms of maintaining and improving the quality of neutronics predictions for the core. In particular, it is of utmost importance to be able to accurately predict the neutron multiplication or reactivity of the system (i.e. its capacity to sustain a chain reaction), as well as possible reactivity variations in the system (i.e. departures from a given multiplication factor). The origin, the inter-relationships and the propagation of the numerical discrepancies between specific code predictions (CASMO-4 and BOXER), as well as their assessment against experiment, of reactivity effect measurements conducted during Phases I and II of the LWR-PROTEUS program have been thoroughly investigated during this research. A reactivity breakdown methodology, developed at PSI, has been extended and applied to shed more light on the sources of these discrepancies. This analytical tool involves a numerical break-down of reactivity effects that allows quantification of the different phase-space contributions (in terms of reaction rate types, energy groups and spatial regions). By comparing the predicted and experimental results, and by analyzing them in detail, useful insights have been obtained regarding the most important sources of error. Further, the research accomplished has helped to identify the fuel lattice characteristics that have the greatest sensitivity to the error propagation process and hence would be the most important for investigations in future experimental campaigns aimed at establishing an optimal validation database. An important part of this work has been devoted to an extensive calculational study of the reaction rate distributions in the SVEA-96+ BWR fuel assembly in an unperturbed situation, i.e. with all pins present in the lattice (the LWR-PROTEUS Phase I reference configuration). The predictions of the reaction rate distributions obtained using the two different deterministic assembly codes (CASMO-4 and BOXER) have been compared in detail. The research has proceeded to investigate reactivity perturbations but in a much simpler system given by a regular lattice of uniform, fresh UO2 rods. This has involved application of the extended reactivity decomposition methodology to a selected number of pin replacements investigated in the LWR-PROTEUS Phase II reference test lattice. In the last part, our attention has been focused on the most complex case, i.e. on the analysis of the pin removal experiments carried out in the highly heterogeneous SVEA-96+ BWR lattice (LWR-PROTEUS Phase I), with the aim of testing and analysing the capabilities of the deterministic codes. In brief, the present research represents a unique extension and application of reactivity decomposition techniques in the context of the LWR-PROTEUS programme, addressing the characteristics of modern LWR fuel. Thereby, it contributes to understanding the physics of the various reactivity changes investigated and suggests the explanation for some of the significant discrepancies observed between calculation and experiment.