Automatically identifying driver inattention could dramatically improve road safety. This paper presents a preliminary study aiming to correlate high levels of frustration with posture information collected from the driver's seat. Using a driving simulator, participants had to drive under normal and frustrating conditions, for example parking in a tight spot with some time constraint. Binary classification using a range of machine learning algorithms provided encouraging results, showing that posture features could help reflect frustration and possibly other drivers' mental states.