The document summarizes a study on the effects of different lighting conditions on facial expression recognition software. It describes how the researchers took video clips of 3 facial expressions (neutral, joy, anger) under 5 lighting scenarios using male and female subjects. The analysis found that lighting from below generally produced the lowest accuracy ratings. Joy expressions were easiest for the software to detect. While some extreme lighting scenarios affected the data, more moderate real-world lighting is unlikely to have as strong an impact.