Dr Yan - the Thatcher illusion

Explanation 1

It's difficult for the brain to interpret expressions in upside-down faces. So we don't really recognise that the expression in an upside-down Thatcherized face has been altered at all.

The problem with this is that it's not really that difficult to work out upside-down expressions. The people in my first line-up have a variety of expressions, and it's easy to tell what they are.

Explanation 2

When looking at an object, our brains often have to work out which way up it is - to orientate it. To do this, we take into account both our personal expectations, independent of the object (eg which way up we are), and the details of the object itself.

For example, if we are sitting up, we expect the top of objects to be at the top of our visual field. If we then see a face with the forehead at the top, both our personal expectation and the orientation of the whole face agree.

Focusing then on a facial feature (say the lips), it's obvious which way up they should be - so a (right way round) Thatcherized face looks very wrong. But when the face is upside-down, our personal orientation and that of the face conflict, which makes it much less obvious which way up the lips should be, and so the Thatcherized face doesn't seem so odd.

Explanation 3

This relies on a well-established theory that there are two sorts of processing we do to recognise objects. One involves looking at the arrangement of features of the whole thing, and comparing this to a known mental map of such objects. The second involves closer inspection of individual, local features; in the case of a face: the nose, the mouth, the eyes, and so on.

For an upside-down face, we’re unlikely to have a known mental map, so we rely on looking at individual features. And the individual features of an upside-down Thatcherized picture look fine. It’s only when we see the face right way round that we can compare it to our idea of a normal face, and the picture looks horrible.

This explanation is backed up by the fact that recognising faces seems to be particularly reliant on the arrangement of facial features, with us gradually getting used to right-way-up faces as we grow up.

Moreover, this sort of recognition has been shown to particularly disrupted by rotation. It also explains the feeling that something separate kicks in as a Thatcherized face is rotated back to upright. But it implies that we can't just mentally rotate an entire face to work out the arrangement of features.

That's a bit odd, as we obviously don't have a problem rotating individual parts of the face: the smile on a normal (non-Thatcherized) face looks fine when the whole face is upside-down.

Just human faces?

It's worth pointing out that the illusion may not be specifically to do with faces, but with things that we are used to seeing one way up. On this note, experiments have been done with upside-down writing, and it's interesting that the effect also works (if more weakly) for pictures of women wearing bikinis [8].

If the upper and lower halves of the bikini are turned upside-down, it’s much less obvious when the entire picture is rotated. Of course, we see faces all the time, and bikinis quite rarely, so it's not surprising the 'bikini illusion' is weaker, but I've been quite convinced with the examples I've seen.

It may not be just humans either. Researchers have presented rhesus monkeys with faces of other, familiar monkeys that have been Thatcherized.

The Thatcherized monkey faces don't look too bad to humans, either way up, and they aren't that distracting to monkeys when shown upside-down, either. But they are distracting when shown right way up - a sort of monkey-Thatcher effect [9].

There have also been tests done on people who can't identify faces (but can recognise facial emotions) - a condition known as prosopagnosia.

Prosopagnosiacs are actually quicker at spotting upside-down Thatcherized faces. Indeed, as you gradually rotate a Thatcherized face until it is upside-down, they don't seem to have that switch that normal people have from instant spotting of the effect, to needing much longer to work it out [10].

There’s also a possibility that the impressiveness of the Thatcher illusion relies on several separate explanations. For example, recent research has highlighted the importance of an effect that we used to judge 3D. Faces are usually lit from above, and when the eyes and lips are upside-down, the reversal of shading makes them look particularly weird [11].

Further research

The explanations above seem a bit convoluted; my instinct is that they aren't phrased in a way that really captures the way the brain actually works - so I don't find them altogether satisfactory. However, it's possible that advances in our understanding of the brain will allow us to look at the neural pathways involved in the Thatcher illusion.

Coupled with recent research on how to get computers to recognise faces, I expect this to produce a much deeper understanding of the Thatcher illusion in the future. In fact, neurological research has already revealed something a little unexpected: it seems as though there are neurons that do detect something wrong with upside-down Thatcherized faces, but we just don’t consciously perceive it [12].

At Bang, we also intend to cover this illusion at bit more, so watch this space.