Ensuring Process safety is critical in process plants. Major accidents such as the ones in Richmond (1993), Longford (1998), and Texas City (2005) highlight that ensuring safety is a continuous battle where the plant operator is at the forefront. Specifically, the human operator in the control room is responsible for effective monitoring and supervision of the process plant. Therefore consideration of human factors especially during abnormal situations is important. To perform his role the control room operator is critically dependent on information conveyed by the Distributed Control System (DCS) through visual displays and alarms [1]. Based on these, during abnormal situations, the operator is expected to quickly and accurately detect when things go wrong, isolate the root cause, and perform corrective / recovery actions. A variety of decision support methods have been proposed to automate monitoring and fault diagnosis [2]. However, to date their deployment and acceptance in actual plant usage is limited. In order to understand why these technologies are unable to penetrate into practice and be deployed for real-time usage, it is necessary to comprehensively understand how the operator uses the control and alarm systems. Specifically, it is critical to understand the cognitive behavior of the operator when he/she is dealing with process upsets and abnormalities. That is the goal of this paper. We seek to develop systematic approaches to understand the cognitive behaviors of process operator during abnormal situations. For the control room operator, the eyes are a critical source of information. An eye tracker is a device for measuring the point of gaze (where one is looking) in real-time. While studies of eye movement are over 150 years old, it is only in recent years that non-invasive, inexpensive, easy to use eye trackers have become widely available. Eye tracker data is now commonly used for deciding information flow on web pages as per gazing behavior of web page user. Kilingar et al., [3] used eye tracker data to analyze pilot's situation awareness. In this research, Tobii TS300 eye tracker is used to locate the point of gaze of process operators as they use the DCS screens to control a process. We also track the mouse movement and key presses on the DCS computers to fully track the operator’s interaction with the DCS (and hence the process). Hence in our study, we have a temporal map of what the operator is looking at, and what actions he performed at a fine granularity. In other words, a spatio-temporal trigger-response chain of the operator’s behavior can be developed. Using this, we seek to infer the operator’s cognitive behavior especially under abnormal situations. As a pilot study, the above mentioned methodology has been tested on 68 undergraduate students, who play the role of plant operators seeking to control a simulated ethanol plant. During the course of experiment, we have simulated various abnormal situations. The subjects are then tasked to bring the plant back into normal condition through suitable intervention. In this paper, we will describe the experimental procedure developed to study the cognitive behavior and also report on some of the key insights on the cognitive behaviors observed during normal and abnormal operations.