This work investigates two important areas in human reliability engineering: models for human performance analysis, and reliable data for these models. In the first area, the author addresses the two modeling techniques "THERP" and "SAINT" showing that the flexibility and nuclear relevance of the second give it the merit to be used in modeling and analyzing human tasks in complex man-machine systems such as nuclear systems. A case study is investigated and simulated by the "SAINT" to explore its dynamic capabilities;In the second area, the author evaluates all the available human data sources, develops a new description format for human related events with all the information matrices needed for human reliability analysis. A complete operations data bank for sixty-five commercial power plants in the U.S.A. over a period of ten years is constructed;Data analysis, with qualitative and quantitative evaluations using the information matrices provided in the operations data bank, is presented. Operator error probability computations are made for some operational tasks in selected nuclear engineered safety feature systems. Moreover, the assessment of the role of the performance shaping factors is discussed.