Even if a system's hardware and software underpinnings are completely reliable, errors at the user interface can cripple or destroy a mission, often with catastrophic consequences. Little attention has been paid to handling faults and errors at the user interface their causes and remediations are little understood and methods of predeployment fault detection in user interfaces are almost nonexistent. The paper presents a working definition of a user interface defect, and a robust method for detecting defects automatically. An experimental methodology for empirical testing and validation is given. Results show that while manifestations of defects may be many, only a few root causes are responsible for them. >