The paper describes the results of the Long-Term Pavement Performance(LTPP) study carried out in Finland and Sweden on the GPS-1 experiment(asphalt concrete on granular base). The main results includenew pavement deterioration models which are based on a failuretime approach using censored data on 64 test sections. A largenumber of independent variables were examined to identify factorswhich explain most of the pavement deterioration on wheel paths(traffic related distress) and on a whole pavement surface (trafficand climate related distress). The most important factors explainingdeterioration included tensile strain at the bottom of the asphaltlayer, or the surface curvature index calculated from FWD measurementsand the freezing index. In addition to deterioration models, anew neural network approach to calculate tensile strains on thebasis of measured deflection bowl and asphalt layer thicknessis introduced.