Current materials and construction specifications for pavement preservation treatments are predominantly prescriptive and they have little or no methodical linkage between initial treatment quality and future performance. There is an imperative need for performance-related specifications (PRS) that link the initial quality of pavement preservation treatments to their long-term performance and life-cycle costs so that rational pay adjustment and acceptance decisions can be made. However, the current literature lacks a methodology for developing PRS for pavement preservation treatments. The aim of this research is to fill this gap in the literature, with focus on thin HMA overlays. In this dissertation, a novel approach was devised for developing performance prediction models for pavements that received preservation treatments. In this approach, the model consists of two tightly-coupled components: the first component is responsible for predicting the performance (e.g., IRI) of the existing pavement if no treatment was applied. The second component is responsible for predicting the reduction in pavement deterioration due to the application of the treatment. Inputs to the first component include material and construction properties of the existing pavement layers, climatic conditions, and traffic factors. Inputs to the second component include the treatment's acceptance quality characteristics (AQCs), climatic conditions, and traffic factors. The artificial neural networks (ANNs) and the Bayesian regression methods were used for developing the two model components. Using this approach, a model was developed for predicting the International Roughness Index (IRI) of flexible pavement treated with thin HMA overlay. The data used for developing and testing this model was obtained from the Long-Term Pavement Performance (LTPP) database. Artificial neural networks (ANNs) and Bayesian regression techniques were employed for developing the first and second components of this model, respectively. A PRS methodology was developed for quantifying the difference between the initial quality levels of as-constructed and as-designed treatments. This methodology consists of a novel approach for determining the probability distributions of service life and present-worth value (PWV). This approach allows for transforming the probabilistic distribution of future IRI (predicted by the Bayesian model) into probability distributions for service life and PWV. Pay factors are then estimated based on the difference between the as-constructed and target PWVs. Finally, this dissertation provides insights into the relationships between initial quality (measured in terms of both mean and standard deviation of key acceptance quality characteristics) and expected pay factors through analysis of real world case studies of asphalt pavements treated with thin HMA overlays. The electronic version of this dissertation is accessible from http://hdl.handle.net/1969.1/151708