May will be a month of illusions
Posted by Dave on May 4, 2010 | 9 Comments
For the next month, I’m going to be blogging about illusions and the people who create them. So let’s get started. Take a look at the following short movie, following the on-screen instructions. Just watch it once!
This one is based on an illusion by Stuart Anstis, who in turn based it on a very popular illusion by Peter Thompson. Let’s take a poll to see whether it worked:
Now I’m going to skip down a few lines to make sure no one sees a spoiler before I explain what was going on.
There! The illusion I showed here is based on a similar one created back in 1980. Peter Thompson, quite by accident, figured out that if you invert the eyes and mouth of a photo of a smiling person, then turn the whole picture upside-down, it’s difficult to notice that anything is wrong with the picture — the smile still looks like a smile. It’s called the “Thatcher illusion” because the first picture Thompson tried it with was then-prime-minister Margaret Thatcher:

(Image source: Michael Bach–follow link for a great demo!) The picture on the top-left has inverted features. An upright version of the same picture is shown below it. To the right are the originals.
In a 2009 retrospective on the Thatcher illusion, Anstis speculated that the effect might be replicable on an entire human figure. The revised illusion, which has been called “Thatcher’s bikini,” is what I was attempting to replicate up above. Here’s the original image and my modified version:

(Image source: The Bikini Open) The image on the right here (upside-down) is the first image I showed in the movie. If we replicated Thatcher’s bikini, then most people shouldn’t say that they saw something wrong with the first picture in the movie.
Anstis says we probably experience the Thatcher illusion because our perceptual system doesn’t process faces as a whole; instead, there may be specialized components for processing mouths, eyes, and so on. So when the Thatcherized face is inverted, we see the upright mouth and eyes but don’t notice that they are upside-down relative to the face, and misread the facial expression.
When Anstis and colleagues repeated the illusion using a neutral face that had been modified with photoshop (instead of actually inverting the mouth and eyes, they curled them up or down to make artificial “smiles” and “frowns”), viewers were actually quite accurate in judging the “real” facial expressions, even for inverted faces. So the effect is not quite as robust as it might seem. It works on the Thatcherized face because when the photo is upside down, we’re seeing an upright mouth and eyes, and judging her expression based on that. When everything is properly inverted, it’s easier to judge the expression.
This month I’m going to be focusing on the art and science of illusions. I’ll be taking close looks at many different illusions and the scientific research behind them. Often you’ll see an illusion online and not realize that a researcher may have spent months or years developing it, so here I’ll try to give credit where credit is due.
Starting on Friday of this week I’ll be in Florida at the annual meeting of the Vision Science Society, where illusionists get together with other vision researchers and trade their best work, culminating in a demo night and an illusion of the year contest. I hope to meet many of the illusionists face to face, and I’ll share my conversations with them right here.
Thompson, P., Anstis, S., Rhodes, G., Jeffery, L., & Valentine, T. (2009). Thompson’s 1980 paper Perception, 38 (6), 921-932 DOI: 10.1068/pmktho
P.S. The image in the header comes from another illusion by Isao Watanabe, Patrick Cavanagh & Stuart Anstis (you may need to shift-reload to see it). It’s called the diamond illusion, and though all the diamonds are the same, each row of diamonds looks darker than the one above it.
Comments
9 Responses to “May will be a month of illusions”
Additional comments powered by BackType



May 4th, 2010 @ 4:50 pm
Interesting take on the thatcher illusion :-)
You wrote:
My understanding has always been not that we process each component separately, but that (at least in newborn infants) the preference is for any top-heavy configuration. So the thatcherized face totally fits those parameters.
Check out Cassia et al. (2004) Psych Sci: http://www.ncbi.nlm.nih.gov/pubmed/15147490
May 4th, 2010 @ 5:15 pm
Jason,
That was a bit of a simplification. We do some of each, but because we do some component processing, the result is the illusion works.
May 5th, 2010 @ 4:21 pm
I was looking for this foto for a long time.
1993 I used it to create awareness of “cross cultural” problems.
We tend to interpret objects only in the known context. Dealing with unknown, new or “unpracticed” contextes eases fatal errors of interpretation.
May 4th, 2010 @ 2:39 pm
May will be a month of illusions http://goo.gl/fb/HVtCd
This comment was originally posted on Twitter
May 4th, 2010 @ 2:47 pm
I’m devoting the month of May on @dailymonthly to visual illusions http://goo.gl/fb/HVtCd
This comment was originally posted on Twitter
May 4th, 2010 @ 3:01 pm
Great stuff! RT@davemunger “I’m devoting the month of May on @dailymonthly to visual illusions http://goo.gl/fb/HVtCd “
This comment was originally posted on Twitter
May 4th, 2010 @ 3:08 pm
May will be a month of illusions http://goo.gl/fb/fQ4PD
This comment was originally posted on Twitter
May 4th, 2010 @ 3:48 pm
Huzzah! RT @davemunger I’m devoting the month of May on @dailymonthly to visual illusions http://bit.ly/aeU4wP
This comment was originally posted on Twitter
May 4th, 2010 @ 7:05 pm
Bikini illusion, seriously, this is #science http://bit.ly/baD29D
This comment was originally posted on Twitter