Categories:

# Variations in Visual Perception

When I took my very first computer vision course, a large part of the curriculum motivated the problem of vision: why do computers have such a hard time with visual tasks? It’s not intuitive that segmenting scenes, recognizing objects, perceiving depth, and reconstructing geometry would be so difficult; after all, humans do it all the time.

To elucidate some of the difficulties, our professor showed us examples of when human visual perception breaks down. You might recognize this common illusion. Which square is darker, A or B?

Checker Square Illusion

It’s a trick question – they are both the same shade of gray. However, you most likely found square A to be darker than square B for a few reasons. The checkerboard pattern provides context; that is, square B appears lighter because it’s surrounded by dark squares. Furthermore, the shadow cast by the cylinder physically darkens the square, but we are able to recognize that and ‘undo’ the light’s effects. This is a pretty important ability to have, otherwise we’d have to reintroduce ourselves every time the sun goes down.

Another common illusion is the Mueller-Lyer illusion. Which horizontal line is longer?

Muller-Lyer Illusion

You probably know it’s a trick question – they are equal in length. But if you’re like me, you perceived the line on the left as a bit longer. With this particular illusion, it’s a little less obvious why this is the case.

One theory is that if you have grown up in an environment rich in ‘carpentered corners’, then this image will evoke the sense of ‘inside’ vs. ‘outside’ at different scales. Despite the fact that your photoreceptors receive the information that they are the same length, some process in your brain that interprets depth overrules that, claiming it knows better.

One Muller-Lyer explanation

The explanation (given here) is that the double arrow provides a depth cue that implies the vertical rod is relatively closer, while the feathered line provides a depth cue that the vertical rod is farther away. If two objects subtend the same angle on your retina and one is supposedly farther away, then that one must be larger – hence the consistent disparity in my (and probably your) perception.

But these depth cues are based on experience – not everyone perceives the left line as longer.  Henrich et al point out that some fundamental human reasoning (visual, spatial, social) widely varies across populations. The San foragers of the Kalahari perceived the lines as being equal, presumably because their environment is not rich in this particular geometry.  From a Bayesian point of view, the visual system computes (estimates?) the posterior uncertainty over some latent representation of the scene given the current visual information and some prior that encodes past experiences.  Under this framework, this research suggests that the San foragers’ prior does not inform their perception the same way my prior does.

So the thinking is that even something as ‘basic’ as visual perception of lengths is influenced a great deal by experience. This actually isn’t too surprising; I’d have a real hard time hitting a 90 mph fastball or creating a photorealistic painting but I can find you a more than capable subset of the population that could (all having experienced that thing Malcolm Gladwell talked about).

This begs the question: to what extent is human perception universal? If evidence suggests that even visual perception of lengths is dramatically affected by experience, then at what point between the retina and the higher-level regions of the brain does perception diverge due to this highly-tuned prior? And does this extend further down the pipeline – at what point is learning to perceive universal? In other words, do the experiences themselves shape the way your prior over experiences is updated, or is the process of learning from experiences invariant across cultures?

References:

1. Henrich, Heine and Norenzayan.  The weirdest people in the world.  Behavioral and Brain Sciences, 2010.
2. Ethan Watters. We Aren’t the World.  Pacific Standard.  Feb 25, 2013.
3. Edward H. Adelson.  Checker Shadow Illusion, 1995.
4. Muller-Lyer Illusion

Posted in Neuroscience, Ramblings, Uncategorized.

## One Response

Stay in touch with the conversation, subscribe to the RSS feed for comments on this post.

1. Cool! The checkerboard illusion seems more compelling than the arrows/lines. I wonder if there are any people for whom the checkerboard illusion fails.