Point-light displays are an amazing demonstration of how the visual system creates order out of what initially seems to be a random pattern. Take a look at this short movie (QuickTime required). Just looking at the first frame, it might be difficult to tell what's being displayed, but after watching for just a second, it all becomes quite clear:
Just these 13 dots, when placed in motion, instantly convey a very clear picture. We can even determine the gender of the person walking or recognize friends, just from displays like this. But now take a look at this movie:
It might take a bit longer, but soon we get the idea.
So we can even recognize "unnatural" movies like an upside-down version of the first movie. Now how about this:
This too is an "upside-down" movie, but this time the "walker" is actually an acrobat walking on his hands. Again, for most people, it's trivial to tell the difference between the upside-down walker and the man walking on his hands. But these videos give rise to an interesting question: what do we recognize first from these displays: the shape of the figure or the type of motion we're seeing?
Thomas Shipley showed movies like these to 23 psychology students, but he made the task more difficult by obscuring the movies with additional distractor dots moving in time with the actual dots representing the walker's joints. There were two levels of distraction: an equal number of distractor dots and real dots (1:1) or twice as many distractors (2:1). In some displays, there was no walking figure at all, just distractors. The students were asked whether they could see a moving figure in each video. Here are the results:
In both cases, the viewers were significantly less accurate in identifying the upside-down pictures compared to right-side-up pictures. Even though the upside-down walking-on-hands picture looked in some ways like the upright walking-on-feet picture, it was much more difficult for the viewers to recognize. Shipley argues that this demonstrates that viewers recognize the dynamics of a display first -- the physical relationship between the motion of the figure and the world. If that motion is unrealistic, it's more difficult to recognize.
So why is the realistic figure walking on his hands recognized later than the upside-down figure walking on his feet? Perhaps simply because people are less familiar with what "realistic" hand-walking looks like.
Shipley points out that this experiment still leaves many aspects of motion perception unanswered. For example, do we recognize more complex figures in motion (such as two people dancing) in the same way? At some point, must we not group together different aspects of a scene in order to understand what's going on?
Whatever the answers to these questions, there's no question that point-light displays are lots of fun. You can see dozens of examples of these displays on Shipley's web page.
Shipley, T.F. (2003). The effect of object and event orientation on perception of biological motion. Psychological Science, 14(4), 377-380. DOI: 10.1111/1467-9280.24471
Actually, I recognized the first figure before it started moving. Too much watching motion capture footage, I guess. I couldn't quite guess what the third was, but I could tell that the top of the figure took some work to balance. That's definitely recognition of motion.
Interestingly I found the upside-down walker very hard to recognise. I had to watch it several times and had I not seen the right-way-up walker video I may never have got it. The hand-walker, on the other hand, popped out on the first viewing.
I couldn't see any of them until i read the description indicating that it was a person walking. I was convinced it was a rotating twig or some abstract shape of some sort. I suppose that means I have a weak prior for people walking in place.
I had the same experience as Philip C:
it took me 4 views before I recognized the upside-down walker, but I saw the hand-walker straight away...
Yeah right! The upside-down walker looked like a rotating twig to me too!
One problem with these examples is that they are low-res internet videos. The actual videos used in the study would have been much higher-resolution. The BioMotion lab (also great fun) has a much smoother display that may give you a sense of what the real thing looks like.
More imputed order out of chaos: The stills look like a mirror image double exposure of Ursa Major. Or am I walking the dog too late at night?
Heh. At first, I thought the hand-walker was an upside down video of someone on monkey bars...
Is there an explanation for why different people take longer or shorter periods to recognize the pattern (assuming there really is a pattern)? And what's the significance of these differences?
I instantly saw the right-side-up walker, took 1 1/2 video plays to see the upside-down walker, and would not have figured out the hand walker without clues from the blog. But the hand walker doesn't look like a rotating twig to me, because I'm watching the bottom half of the screen so intently I don't analyze the motion on the top half. Why? And would a tendency to focus on one part of the image make it easier or more difficult to identify the pattern?
Had the same experience: the walker was easy, the walking on hands only sightly longer, and the upside down walker almost impossible. I constructed it as a rotating branch, like some others mentioned. Here's a related illusion, a kind of dynamic vase/faces type thing:
I figure that as I started to resolve the dots into the revolving-twig action, I was increasingly prevented from the possibility of seeing it as the upside-down walker.
I also thought of them as Ursa major or the Sagittarius, before finally being able to 'see', as they got moving. Thats what I call a cakewalk!