The core objective of this research is to provide a methodology for obtaining guidance based on empirical evidences on pavement marking replacement times. This dissertation investigates the degradation process of pavement marking visibility over time in the United States using semi-parametric family of duration models. Specifically, a methodological framework to analyze typical pavement making visibility inspection data was formulated. The National Transportation Product Evaluation Program datasets pertaining to water based paints from a total of nine testing locations in the states of Alabama, Pennsylvania, Mississippi, Minnesota, Texas, and Wisconsin were used for the purpose of this investigation. From a methodological standpoint this research suggests that mid-point imputation is reasonable to approximate interval level failure data. Furthermore, the elapsed time model seemed to exploit the empirical pattern of event dependence among multiple marking samples on an experimental deck better than the gap-time model. This suggests that event dependence exists and degradation of the pavement marking visibility is more simultaneous than sequential.