Abstract. We describe a model to predict the minimum resolvable temperature ͑MRT͒ performance of thermal imagers. Although MRT is a common measurement, it is difficult to achieve consistent results. The operator is permitted but not mandated to change gain, level, and sample phasing for each bar pattern viewed. Changing the sensor control settings affects the resulting MRT. However, the state of the imager is not recorded along with the temperature data. The model predicts the effect of gain, level, and sample phase on MRT results. Model predictions are compared to measurements to demonstrate theory validity. The comparisons also demonstrate the variability that results from ignoring the sensor state when reporting MRT. © 2009 Society of Photo-Optical Instrumentation Engineers.