Traditional empirical studies to estimate probability of detection (or POD) are expensive and time consuming. Over the past thirty years, much progress has been made in the use of physics-based models to predict POD. A deterministic model for flaw response can be combined with a probability distribution for inspection variabilities to provide a model-based POD. Actual inspections, however, involve complicated variabilities from a variety of sources and modeling all of the important ones, and especially human factors variabilities, would be difficult or impossible. Bruce Thompson's knowledge of physics, probability, statistics and industry needs gave him the insights to pioneer and subsequently serve as the leader in the important area that is now called "Model Assisted POD" or MAPOD. The basic idea of MAPOD is to find an appropriate combination of a physics-based model, combined with limited (usually by time and cost constraints) experimental data and statistical modeling to establish POD. This talk will outline Bruce Thompson's important contributions to this area.