||The design of a geotechnical work should satisfy both the ultimate limit state (ULS) and serviceability limit state (SLS) requirements. The limit state design approach addresses various performance requirements and aims to accommodate uncertainties in design. Presently, most limit state design codes use reliability principles for ultimate limit states. However, serviceability limit states are still considered using the conventional deterministic approach. Many geotechnical works such as building foundations are more often governed by serviceability requirements rather than by ultimate limit requirements. Therefore, the deterministic displacement criterion may not be sufficient and a reliability-based displacement criterion is needed to enhance these design codes. Pile foundations are widely used in practice for supporting superstructures. A number of methods are available for estimating settlement of single piles and pile groups. Due to the presence of uncertainties, there may be errors with the estimations. To develop reliability-based serviceability criteria for pile foundation design, it is necessary to characterize the performance of commonly used settlement prediction models for single piles and pile groups. In this thesis, performance data of 64 cases of single piles and 59 cases of pile groups from field static loading tests are collected for study. These data are compiled into two databases, Database A for single piles and Database B for pile groups, for calibrating settlement analysis models for single piles and pile groups. Driven pile foundations in sandy soils are the main focus of study; other types of pile foundations are beyond the scope of this thesis. The model error is represented by a bias factor, which is the ratio of the measured displacement to the estimated displacement. For single piles, the model bias is studied through moment and regression analyses against field data. Based on the obtained statistics of the model bias, the probability distributions of the model bias are established. Besides, the sensitivity of single pile settlement to soil parameters is conducted to identify the most sensitive soil parameters that should be better characterized by more detailed soil investigations. For pile groups, the model bias is studied through a moment analysis against field data, considering the sensitivity to soil modulus. Based on the obtained statistics of the model bias, the serviceability reliability of pile groups with a particular design analysis model is evaluated. Finally, the relationship between the pile group settlement and the single pile settlement, termed as group settlement ratio, is studied and the pile system settlement is explored.