Ok bahama mama i think i've got it...explaining it may be a bit trickier!
First off, we are going to divide the screen into zones.
In zone 1, we have the left most column (3 people) and the one next to it on the right (2) people. That's a total of 5 people.
In zone 2, we have the middle column at the start, which has only 2 people.
In zone 3, we will have the right two columns, the first of which has 3 people in it, the last on the right having two.
This is all at the start, so hopefully you are with me. Here is a screenshot, with M standing for a person.
M_M____M___M_M
M_M____M___M_M
M__________)M
z1_____z2____z3
Now, here is what happens. The top halves of the people in zone 1 move onto the top halves of people in zone 3. But there is nothing special about that, because zone 3 starts with five people, and ends up with five people. The next move is the trick.
In the second movement, the top halfs of the people in zone 2 and 3 (7 total people...see above pic) are only placed on the bottom halves of 6 people from zone 1 and 2. The man on the bottom left is the key. On the first move, he loses a slice of his head when the tops of zone 1 move onto zone 3. But when the tops of zones 2 and 3 move onto 1 and 2, nothing ever replaces on top of him. If you watch it after reading this and look for these things you should see it.
Does that make any sense whatsoever?