The human brain is very good at filling in gaps and seeing things where only a few clues are apparent, as it simply cannot receive and decode ALL the information it receives to give a true picture of a scene. We see selective bits and the rest is filled in for us. We’re also particularly attuned to seeing or hearing things related to humans – a whistling wind, for example, might sound like whispering, and grain patterns in wood can easily take on the appearance of a face. Our recognition ‘software’ for this is enormously strong but can work incorrectly at times. I would think that most of the time the ‘ghost’ is really a moment of many factors – light levels, light direction, the arrangement of objects in the area etc coming together in a convincing enough manner that the brain says, oh, a person! and fills in the gap so that we believe we see them.