This thesis investigates the relationship between predicted and observed fatigue life of asphalt. This study also investigates the positive effects of modifying bitumen with Retona, a bitumen modifier produced in Indonesia from natural asphalt rock sources, on pavement performance in terms of increased resistance to fatigue and permanent deformation. Classical pavement fatigue analysis assumes cracking to initiate from the bottom of the pavement and propagates to the top. However, when relating ‘predicted’ pavement fatigue life to ‘observed’ fatigue life, one immediately encounters a ‘conflict’ with theory because in practice cracks are also initiated at the pavement surface. This thesis also attempts to explain the phenomenon of surface or top-down cracking. In the past, many studies were made to validate design procedures by matching predicted performance with field performance. However, this project, only takes into account the studies that were carried out in the Netherlands and by Dutch researchers elsewhere. Test results performed in the 1990’s on three pavement sections on the accelerated pavement testing facility Lintrack, then owned by the Section of Road and Railway Engineering, Department of Civil Engineering Delft University of Technology, have therefore been used in this study. The Lintrack research provided a large amount of valuable data with respect to pavement performance and was therefore perfectly suited for a study to match theory with practice. To achieve these goals, several steps were taken. Firstly, the data obtained from observations made on the Lintrack accelerated pavement test sections (built in 1990) were studied. The sections were simple two-layered pavement systems consisting of a gravel asphalt concrete (GAC) layer overlying a sand subgrade. The data consisted of information on the geometry and material characteristics of the test pavements, loading and environmental conditions, deflection test results and visual condition data in terms of cracking and permanent deformation. Secondly, a fatigue cracking prediction model was developed based on laboratory data obtained from four point bending (4PB) tests on the same material type. To more accurately simulate the fatigue behaviour of a real pavement in the laboratory, a new test setup termed as the “beam on elastic foundation” (BOEF) test was developed. Since the GAC pavement sections built in 1990 had been removed, the GAC (GAC 1990) material was not available anymore. Therefore, a new GAC (GAC 2010) mixture needed to be produced and considerable efforts were made to produce the GAC 2010 mixture such that it truly replicated the GAC 1990 mixture. Material characterization tests performed to understand better the GAC 2010 mixture included, in addition to the two mentioned fatigue tests; monotonic uniaxial tension and compression tests, indirect tension tests, mastic healing tests and tests on the recovered bitumen such as penetration and Dynamic Shear Rheometer tests. The analysis of the Lintrack APT sections was carried out by determining; (1) the pavement life based on the back calculated modulus of the asphalt layer for different probability of survival levels, and (2) the magnitude of damage that was initiated at the bottom of the asphalt layer expressed by means of Miner’s damage ratio. The cumulative damage ratio, ni/Nfi (Miner’s ratio), was calculated based on the tensile strain at the bottom of the asphalt layer at different temperatures that occurred during the Lintrack tests and the fatigue relationships that were obtained from 4PB and BOEF fatigue laboratory tests. It is shown that the observed pavement life based on the back-calculated asphalt modulus from deflection measurements is longer than the pavement life calculated on the basis of damage initiation at the bottom of the asphalt layer. The results showed that for all three Lintrack sections, the BOEF based predictions exhibited a better agreement as evidenced from the smaller shift factor between “field stiffness reduction” lifetime and the lifetime based on “fatigue” predictions. Therefore, BOEF test based fatigue models are highly recommended to be used for predicting pavement life in practice. This study has shown that it is very difficult to relate cracking visible at the surface of the pavement to the initiation of fatigue damage at the bottom of the asphalt layer. Finite element simulations using detailed tyre–pavement contact pressure modelling have been carried out. The results showed that significant tensile strains had developed at the pavement surface. The magnitude of these tensile strains was such that they can be held responsible for the development of surface cracking. It should be noted however that surface cracking cannot be explained using a stress based analysis. This study clearly showed that permanent deformation (in this study this was permanent deformation of the subgrade) exhibits significant effects on the formation of longitudinal cracks at the edges of the wheel paths. The study on modifying GAC with Retona showed that the GAC+Retona mixture exhibits better mechanical properties compared to the reference GAC mixture. The Retona modified mixture showed a higher fatigue life and higher resistance to permanent deformation compared to the reference GAC mixture.