Until now, various types of threshold and signal-detection models have been proposed. These models have broadly different assumptions on recognition memory mechanisms, and thus predict different patterns of Receiver Operating Characteristics (ROCs). Assessing the models for their flexibility and goodness of fit to empirical ROC data provides evidence on which model has more concise description on the phenomenon of recognition memory and may provide a better understanding of it. I have illustrated the structures and properties of various recognition memory models with their relation to a dual process theory in mind. Then, I have summarized the basics and recent developments of model selection methods. Further, I have explained the pragmatic use of these measurement models, using several examples of empirical studies on different research topics.