Cognitive Daily

Where is the light coming from?

Today’s reading is “Prior Knowledge on the Illlumination Position” by Pascal Mamassian and Ross Goutcher of the University of Glasgow (Cognition, 2001 [PDF link]).

When we see an embossed seal such as a notary stamp, how do we know which parts are convex (bumps) and which are concave (dimples)? When we look at such a seal through a microscope (or even a toilet paper roll), so that we don’t know where the illumination is coming from, we can’t tell what’s up and what’s down. This effect was first recorded in 1744, and first accurately explained in 1786 by David Rittenhouse: we assume the light is coming from over our heads; if the light isn’t coming from where we expect, then we reverse the image. This makes sense: most of the time in the natural world, that’s where the light is coming from. This preference doesn’t appear to be learned: baby chicks raised with light coming only from below still behave as if the light source is above them.

Over the course of the twentieth century, some of the researchers investigating this problem noticed that the observers in their experiments not only preferred the light source to be above their heads; they expected it to be slightly to their left as well. This phenomenon made no sense: people don’t tend to orient themselves with the sun to their left. The scientists were so certain that this finding was in error that they simply discarded results from observers who favored light from the left.

Mamassian and Goutcher suspected that we may really have a preference for light sources on the left, so they devised an experiment to test it. Consider the following two objects:


If we imagine we are in a room with very low light, coming from above, and we’re looking at the objects head-on, they will look like this:


Notice that if you turn the image on the left upside down, you see that it’s actually the identical image. What Mamassian and Goutcher did is take this figure and rotate it in increments of 15 degrees. Then they asked observers whether it looked like the wide bars or the narrow bars were projecting out from the figure. In fact, observers were looking at the same image each time, so their responses would reflect only where they perceived the light source for the image to be.

They found that observers were most consistent in their responses when the light appeared to be coming from about 25 degrees to the left of vertical. It didn’t matter whether observers were left- or right-handed; all observers had the same preference.

Mamassian and Goutcher suggest that a possible explanation for this result might be a “visual field bias.” When recognizing at people’s faces, for example, we tend to pay more attention to the right-hand side than the left. If the light source is on the observer’s left, then the right side of the person they’re looking at is illuminated.


  1. #1 Kevin Saff
    March 2, 2005

    Interesting! Surely, other possible sources of this bias cannot be ruled out from this experiment? Graphical computer interfaces usually show light coming from the upper-left. Could exposure to computers influence the leftward bias, or does the bias influence graphical design?

    Also, since we read from left to right, perhaps this is what causes us to prefer light to go from left to right. Perhaps readers of right to left languages would prefer light sources on their right sides?

  2. #2 Dave Munger
    March 2, 2005

    Good points. However, the effect was observed (and discarded) before the widespread prevalence of computers. I’m no expert, but as you suggest, I suspect the reason is more likely the reverse: our preference leads to the common design practice.

    Mamassian and Goutcher’s visual field bias explanation should be true regardless of cultural preferences; this would be an interesting test of their explanation. They also suggest running the experiment on patients with injuries to their right parietal lobe, who tend to neglect the left visual field—this would be another way to test their explanation.

New comments have been disabled.