Composite structures are very weak through the thickness and are therefore prone to delamination failures at the interfaces between plies. Cyclic loading of these structures can lead to delamination onset and growth, and the ability to accurately model the evolution of this type of damage is critical to reducing the large costs associated with experimental fatigue testing of composites. Many predictive models focus on the analysis of only the propagation phase of fatigue life and neglect the initiation phase, leading to conservative estimates of overall fatigue life. Those models that do consider both phases, tend to treat them separately; however, this research focuses on the unification of these two phases under an assumption that the propagation of delamination in a carbon-fibre reinforced plastic can be treated as a series of stepwise initiations. The work has involved the generation of S-N curves for the initiation of a delamination in a uni-directional composite laminate subjected to pure interlaminar tension. The shallow slopes found in these tests have highlighted the importance of accounting for the initiation phase of fatigue life under mode I loading. A damage model was then fitted to these curves to obtain two fatigue-related material parameters, and the model was then used in an attempt to predict mode I fatigue delamination growth (FDG) rates in the same material system. It was found that the model accurately captured FDG from initiation data, but only so long as any increasing resistance to FDG during the tests was ignored. Through a series of novel mode I delamination growth tests, it was shown that fracture toughness increased during the application of displacement-controlled fatigue loading as a function of the number of cycles applied. This implied a constantly changing process zone and an increasing resistance to further FDG, verified by the observation of a larger number of bridging fibres in the wake of a fatigue crack tip, compared to static growth. Normalising by this increasing fracture toughness, produced consistent Paris curves with reduced slopes and thresholds. The model was adapted to account for these increasing bridging forces at the crack tip and, by considering the initial severity of each test, was able to accurately reproduce FDG rates for all tests. Although it is hoped that linking initiation to propagation might reduce fatigue testing requirements for designers of composite structures for aerospace applications, this work has highlighted that the resistance to FDG in uni-directional laminates is dependent on loading history, therefore potentially complicating the modelling of such growth.