Evolving solutions for machine vision applications has gained more popularity in the recent years. One area is evolving programs by Genetic Programming (GP) for motion detection, which is a fundamental component of most vision systems. Despite the good performance, this approach is not widely accepted by mainstream vision application developers. One of the reasons is that these GP generated programs are often difficult to interpret by humans. This study analyzes the reasons behind the good performance and shows that the behaviors of these evolved motion detectors can be explained. Their capabilities of ignoring uninteresting motions, differentiating fast motions from slow motions, identifying genuine motions from moving background and handling noises are not random. On simplified problems we can reveal the behaviors of these programs. By understanding the evolved detectors, we can consider evolution as a good approach for creating motion detection modules.