Take a look at these three pictures. Can you tell which is a human, which is a cat, and which is a pigeon?
How about these three pictures? A little easier?
It would be even easier, of course, if the pictures were in motion, like in this movie (QuickTime required). (Update: Nikolaus Troje has provided a link to a full set of demos here -- check them out, they're awesome!) The difference between the first three pictures and the second three is that the points that make up each animal's body parts have been scrambled in the first set of pictures. Why? Because the researchers Dorita Chang and Nikolaus Troje wanted to understand why point-light displays, when in motion, give us such a clear picture of the organisms they represent.
When viewers see the unscrambled pictures, they readily discern whether the point-light display represents a living thing or a random moving pattern. In fact, the task is so easy that it's not actually very useful for researchers trying to understand the visual system. What Chang and Troje want to know is whether viewers use a "local" system or a "global" system to identify biological motion. In other words, are viewers looking at an isolated part of the display like the human's ankles, or are they considering the concerted motion of all the points together?
They showed twelve volunteers movies of either scrambled or non-scrambled humans, cats, or pigeons. To make the task more difficult, the movies were obscured by a constantly changing random pattern of 125 stationary dots. Half of the movies were upside-down, and half were right-side-up. Each movie began at a random point in the walker's gait, and ran for 1/5 of a second to a full second. Then the viewers rated each movie for how confident they were that it contained a living creature. Here are the results:
As you might expect, the upright animals were significantly more likely to be recognized than the inverted animals, and coherent was more recognizable than scrambled. But even the scrambled figures were better-recognized when upright than inverted, suggesting that we don't simply have a global pattern for recognizing biological motion. Perhaps we do focus on a particular part of the scene. In a second experiment, the same pattern of results was found for determining the direction an animal was walking.
But Chang and Troje's third experiment was the critical one. In this study, viewers rated each movie both for whether it represented biological motion and which direction the animals were walking. Here are those results:
The asterisks on the figure represent the average results for all viewers. The individual points show the averages for individual viewers. The x-axis charts accuracy on the direction task, while the y-axis charts ratings on the biological motion task. In every case, viewers who were better/more accurate on one task tended to be better/more accurate on the other task. In other words, accuracy judging direction was correlated significantly with accuracy judging whether the movie depicted biological motion.
The researchers say that this demonstrates that viewers cannot be using global information to determine whether a motion is biological. Other research has found that the motion of the ankle appears to be a key in identifying biological motion. This may be because nearly all walking vertebrates swing their legs forward in a similar manner: they don't actually use their muscles, but instead simply rely on gravity, thus conserving energy. Chang and Troje speculate that perhaps it is this distinctive arc that viewers focus in on when they identify biological motion.
Chang, D.H., Troje, N.F. (2008). Perception of animacy and direction from local biological motion signals. Journal of Vision, 8(5), 1-10.
Oh man, I was a participant in this study in my undergrad! Is that a violation of ethics for me to say that?
Looking at that graph, I feel so predictable now.
I can't see the QTs from work, so could someone please post the answers? For the record, I went for Cat, Human, Pigeon, and I used my tried and tested technique (for 'blocky' or otherwise obscured images) of removing my glasses and letting my brain try to make sense of the blur.
But even the scrambled figures were better-recognized when upright than inverted, suggesting that we don't simply have a global pattern for recognizing biological motion.
I don't see how the finding suggests that.
For an interactive Flash movie which contains examples of all the stimuli used in this study, please visit our webpage: http://www.biomotionlab.ca/Demos/scrambled.html