In order to operate NPP safety and reliably, not only improvement of reliability of machine systems but also that of human operators is required. From this point of view, a training support system for diagnosing plant anomaly skill of operators has been developed. The system automatically detects viewpoints and verbal protocols of a trainee in real time when he/she practice to diagnose a root cause of the plant anomaly simulated by a plant simulator, and they are presented to an instructor in order to give the trainee effective instructions based on his/her thinking process. By using the system, an experiment has been conducted to evaluate the effectiveness of this approach. As the result, it was found that the two information, trainees¡Çviewpoints and their verbal protocols, are effective to trace their thinking process and to estimate their lack of knowledge and misunderstandings.