Pattern Recognition and PCA

There have been two interesting recent articles in the popular press and social media feeds. In the first, it was found that pigeons can be trained to be remarkably good at detecting the characteristic microscopic appearance of breast cancer as compared to the normal look of breast tissue. In the second, it was pointed out that one reason why so many of us don’t like how we look in photographs is because the image is the reverse of the familiar face in the mirror, a subtle version of having the right and left hands switched on a statue – though perhaps this is less of an issue in this age of selfies.

Another thought: rendering the human appearance realistically is one of the greatest technical challenges for animation and animatronics. Intriguingly, we are most disturbed when the simulation is almost but not quite there, the so-called “uncanny valley” effect. Familiar but “off” makes a lot of people uneasy.

These phenomena are really about pattern recognition, the ability to “stand back” from a complex and detailed picture and get an instant read on whether it is symmetrical, true, realistic, and whether it conforms to a known standard of some sort or at least has enough features in common with one so as to be recognized as fitting it. Unless there are serious neurological issues, most of us are very, very good at this, and it underlies a lot of the things we do every day without a second’s conscious thought. Sometimes, of course, we can all get hung up on details, and don’t stand back and see the “big picture” – giving us the popular expression, in English, of “not seeing the wood for the trees”.

Here is the image again thumbnail size.  A bit of distance often helps us see.

Regular readers will see where I am going with this: personal colour analysis directly hooks into this natural capacity for spotting the point where it all makes most visual sense, the place where the eye can repose, the most natural effect, and with the least visual distortion in the form of negative optical illusions. One of the things an analyst has to do during training is to switch off any learnt editorial commentary about colour and colouring and to relax enough to unlock this natural and easy recognition of what makes most harmonious visual sense on the person before them.

PCA originally focused on four broad “seasonal” groups, which became known as spring, autumn, summer and winter, but which could just as easily have been called something like “daybreak, sunset, evening and night”, or “North, South, East and West”, or perhaps, less picturesque but more useful, “bright warm, muted warm, muted cool and bright cool”. Generalizations arose attached to these tonal ideas – blondes with green eyes were said to be likely to be springs, say, and brunettes with brown or hazel eyes were thought candidates for autumn, and so on. These rules of thumb were a sort of seasonal “profiling” – an attempt to identify groups of characteristics which might help us begin to sift through the possibilities and point the client in the right direction.

As the practice became more refined it became clear that the overall picture was more complicated and that these simple clichés weren’t necessarily true. Terms like “blonde” and “brunette” can cover a lot of tonal ground, and then there was also the fact that these categories were based on a particular sample of the population, and didn’t consider that most of the world isn’t of north-western European ancestry. (If you had black hair and brown eyes – and worldwide, the vast majority of humans by far could be said to meet this criteria – then in some early PCA reckonings you were automatically a winter, never mind that “black hair” and “brown eyes” are as myriad in their variations as “blonde” and “blue”, and that’s before we consider the nuanced palette that is human skin.)

Further, the original “four-hand” system did not account for the concept of neutrality (that is, skin tones neither completely warm nor cool). Above all, the fact remained that human colouring, like all things human, shows us subtle, intriguing and unexpected permutations. For all that, these broad classifications and general principles arose in the first place because they *did* seem to have some rough utility in the client base these early analysts were seeing – natural blondes usually aren’t winters, nor redheads cool seasons; common things tend to occur commonly and just about everyone at some point employs the shorthand of archetypes to help simplify a complex world.

In TCI 12-Tone practice, we assume nothing. We test the effects of colour, asking the drapes to tell us where the client sits in terms of best hue temperature, chroma and value. That said, while the starting assumption is of all bets being off, and that anyone could turn out to be anything, it is still true that “common things occur commonly”. When you get an apparent exception, there are usually other factors at work – illness, undeclared hair dye, and so on. If someone looks like a Light Spring or Summer, they usually are (occasionally they turn out to be a Soft, but it’d be a rare, rare light-looking person indeed that turns out to be a Dark in the client base I see).

During the course of a draping a client may “connect” with individual drapes that deviate in some way from their final tonal home (see the previous article on “The Leaning Toner” for a fuller discussion of why this may happen), but the analyst is always looking for overall harmony, the place where the client belongs most naturally and without strain or being “worn” by the colour, without the dissonance of being notably lighter or darker or brighter or duller or warmer or cooler than the “foreground” colours we are giving them, and while we may focus on details to clarify the effect – shadows under the eyes or whatever – it is no less important to “stand back” and take in the overall effect, the gestalt.