Being able to predict how internet users react when confronted with a potentially dangerous call for action in an online message (such as an e-mail) is important for severalreasons. On the one hand, users have to be protected from fraudulent e-mails such as phishing. On the other hand, over-cautious users would be difficult to communicate with on the internet, so senders of legitimate messages have to know how to convince recipients of the authenticity of their messages. Extensive research already exists from both of these perspectives, but each study only explores certain aspects of the complex system of factors influencing users’ reactions. In this paper the results of our efforts to integrate the various existing findings into one comprehensive model are presented, along with the results of a preliminary qualitative evaluation of some of the model’s predictions using quantitative as well as qualitative measures and eye-tracking.