Human beings are mostly quite good at recognising faces. We seem to have evolved this skill as part of becoming a large-scale social animals, and we share it with only a few species – as far as we know, anyhow.
But of course like anything there is a spectrum of ability. A small number of people are “super recognizers”, able to pick out a face they have only seen once before amongst a sea of other faces. Most of us are close to the middle. And some people aren’t very at all good when it comes to recognising others. This can be hereditary, or developmental. Others may have suffered a brain injury, or a stroke. In fact if you suddenly find you can’t recognize people that you could minutes or hours before it should be regarded as a medical emergency.
People who have difficulty with facial recognition are described as having prosopagnosia or face blindness. A famous example, whose condition was unusually pronounced, was “The man who mistook his wife for a hat”, described by Oliver Sacks in a book of that name. Such cases are very rare.
But most of us have had the experience at one time or another of meeting someone we know, but in an unfamiliar setting and not being able to quite place who they are. Perhaps prosopagnosia is like that but occurs with everyone you meet.
There are related conditions, collectively called agnosia. Some people experience difficulty in recognising objects for example, or places. Curiously there can be agnosia for cars, and although I am at least average for faces and places, cars do all look (more or less) the same to me. I think this should be called ‘harmagnosia’ but I just coined it from the Greek for ‘chariot’ + agnosia. Remember, you read it here first.
The neuroscience of face recognition is fascinating. It’s thought that there are specific parts of the brain specialised for facial recognition, and that these play a part in what is often called the “Margaret Thatcher Illusion”. (This is more of an effect than an illusion really, so the Margaret Thatcher Effect is a better description.) It was first published by Peter Thompson in 1980. The effect is illustrated below.
Above are two pictures of a famous person. Although both are inverted, I’ve also doctored one of them. What have I done, can you tell?
After a little time you can probably spot that before inverting the whole image I turned the eyes and mouth of the right-hand face upside-down.
But it’s easy to recognise Einstein in both pictures and the right-hand image doesn’t look too bad at all, a little odd perhaps but on the whole Einstein appears his normal agreeable and avuncular self.
But now let’s flip the pictures the right way up.
They are exactly the same images as before, merely flipped. But now the right-hand image appears grotesque and disturbing, in a way that simply wasn’t apparent at all before. Look back at the earlier picture and you’ll see what I mean.
The startling effect above is generally called the “Margaret Thatcher Illusion”, although it’s not exactly an illusion, and effect might be a better term. The process applied to the images has come to be known as “thatcherizing” them.
Various attempts have been made to explain the effect. Some involve the specialised ability of primates, including humans, to recognise faces. It’s theorized that when the image is the right way up we view it holistically — it’s a face, taken in all at once – but when it’s inverted that is broken up, and we see the individual features in isolation, and because they are actually the right way up they don’t look too bad.
It’s cannot be merely that we have parts we expect to be a certain way up, which taken together make up a whole that also has a familiar orientation. If we invert Australia on a map of the world, then invert the whole map, it’s very easy to spot that Australia is not the right way. Even an image clock-face with the numeral 12 upside-down doesn’t display the Thatcher effect when it’s turned upside-down.
And in spite of the idea that facial recognition is disrupted by inversion and/or distortion, this isn’t true. If you were able recognize Albert Einstein the way up you expected him to be (which you might not of course), then I think you will have recognized him in all four versions. So there is a lot to explain here. I wonder what the necessary conditions are for this kind of effect.
I’ve seen examples involving cartoon characters which do display the effect (but it doesn’t work with something as simple as a smiley, although it’s not clear why) and I wondered if it would also work with images of animals, especially of a familiar species we think of as having faces. So I tried with a cat picture, and sure enough the cat-thatcher effect exists. Both images have been thatcherized but the upside-down cat on the left looks like quite a friendly animal (more so than the original cat which was a “grumpy cat” picture).
But when it’s the right way up the identical image appears wild-eyed with an anatomically challenged mouth. It looks really odd now, but when it was the other way up the distortions were hardly noticeable.
I also wondered if the rest of the face was important, or just the eyes and mouth. Here is an example in which I have taken just two disembodied eyes and a mouth and thatcherized them. Even without the rest of the face the smile on the left may seem familiar.
More disturbingly in some ways the eyes and the mouth don’t have to be in the proper relation to one another, as the next example shows
Even though the eyes and mouth shouldn’t be in a horizontal line, arranging them like that doesn’t look too bad: we readily see it as a face and the image on the left looks quite agreeable, whereas that on the right looks quite odd, perhaps the face of someone who feels suspicious… And yet a hint of that famous smile keeps breaking through, probably because when the features care all in a line we become uncertain which way is “up”. Sometimes we can mentally turn the mouth the right way up again and then it flips back. It’s most strange.
That’s what I see. But it might be different for you, and you might not be able to recognise the mouth. Not everyone will.
A last thought. How does a carnival mask work? Does it, in fact? Is there a minimum amount of face required for recognition, and if so can it be measured?
Wikimedia for the other images