Think-aloud is a de facto standard in user-based usability evaluation to verbalize what a user is experiencing. Despite its qualities, it has been argued that thinking aloud affects the task solving process. This paper reports from an empirical study of the effect of three think-aloud protocols on the identified usability problems. The three protocols were traditional, active listening and coaching. The study involved 43 test subjects distributed on the three think-aloud conditions and a silent control condition in a between-subject design. The results show that the three think-aloud protocols facilitated identification of the double number of usability problems compared to the silent condition, while the problems identified by the three think-aloud protocol were comparable. Our results do not support the common emphasis on the Coaching protocol, while we have seen that the Traditional protocol performs surprisingly well.