Tuesday, May 11, 2010

A Musing

A couple of posts ago I pondered why the small thumbnail I had at the top of the post looked human at all, given that it was just a series of sketched lines (I've included it again for good measure).  I stated that it was now possible to tell that the figure was in her birthday suit, and that this might offend people. So the question remains, why does it look human at all?  How can a series of black lines with minimal light and shade represent so definitively a female form?

It seems to make no sense.  Art in this form makes no sense, it shouldn't work!  There is no situation in nature where we see something made up purely of defining lines; we always see it made up of light and shade, and from this we can deduce shape and form.  Being able to recognise something in that way makes perfect sense as a survival trait.  Mr. Caveman is wandering along looking for lunch and spies a saber-tooth cat to his left.  He knows it's a saber-tooth cat because the light bounces off some bits and is occluded in the way of shadows in other bits.  "Golly" he says, "It's a big toothy thing come to chow down on my succulent man flesh!"  Or, more likely, has says "UG!" defecates (not in his underoos, as he has none), and runs away screaming.

So we have a method of conveying information that is completely useless to us as a species.  Can other species relate to the same thing?  I don't know.  I gather dogs can recognise photographs, which makes sense since they see it as light and shade the same way we do, and the same way they see the world; but can they determine a line drawing of their owner that a human would have no problem with?  Possibly it would have to be a picture of the owner where the nostrils and chin played a far larger part than is usual, given the angle dogs see the world from, and to be honest I have no idea if they would go "Oh, it's my master," or "Why is this person showing me squiggles?"  As far as I can tell no one has bothered to find out in any scientifically repeatable way.  I suspect the Dog would be able to make it out, at least simple and relevant forms, due to what I believe these simple like drawings represent; silhouettes.

Going back to our caveman (We'll call him "Ngh") he gets back to his cave and decides to invent art (It's gotta be easier than that whole wheel thing right?).  He grabs some mud and uses it to paint a silhouette of the Saber-Tooth on the wall.  He knows the silhouette because that is a useful survival skill: He looks up from his latest kill (squirming in the dirt before him; buying online is far easier Ngh, you should check it out), and stark against the sky is a ruddy great cat about to jump down and eat him.  There are no details to be seen since it is  wholly in shadow, but he still knows it's a cat, and most likely one with teeth as big as his arm. So it's a shape you can reproduce.  The shape is nearly as recognisable if you just draw the outline, and if you stack a lot of these outlines together you end up with a line drawing.  That's my theory anyway.

Some research has been done into this sort of thing, mostly because clever boffin types would like computers to be as good at shape recognition as people.  A lot of that research pertains to facial recognition, which seems to be entirely it's own subject, slightly separate from the task of comprehending everything else (maybe I'll cover that another day); but a reasonable amount is relevant to this discussion.

One theory has it that humans have a store of basic primitive shapes that when joined together form a different whole.  This list of primitives would go further than just Cylinder, Cube, Pyramid (The paper used the term Geons), but would form the basic building blocks for the visual language of the brain in much the same way as letters form the building blocks of readable communication for the literate, or musical notation for the aural.  Unlike written or musical communication however, these dimensional symbols would be present almost from birth, requiring less learning than the more structured symbols of other communication (At least I think that's what it was getting at, it was written in Sciencese and was therefore hard for my tiny mind to grasp).  It also went into some aspects of how the brain processes these shapes from line drawings, and what visual cues are required to interpret them.   For example:
Those are actually the exact same image, except that for the one on the left I've removed all the concave (outward curving) lines, and left only the convex lines and intersections. On the right the reverse is true, and only the concave lines are visible, with almost no intersections.  Which one is most easily readable as a human figure (To be fair seeing them side by side messes up the experiment a little, but never mind)?
To me it's the one on the right, which is surprising as you'd think that the intersections of the forms would be pretty vital to reading dimensionality, but apparently this is not the case.  Both forms are readable, but the one on the right more intuitively.  Here's another view of the same image, this time seen as I drew it, and in negative:
As you can see they're both perfectly readable as a human form (despite my shortcomings as an artist), but this time the negative image on the right is marginally harder to read than the positive image on the left.

Why am I interested in this sort of thing? Mostly because when you have a very short time to relay some information via an image (such as 60 seconds to draw a figure) then it helps if you can get down the most relevant information first. As it happens I'm unlikely to draw all the convex lines before anything else, but it's nice to be aware that's it's a stylistic option.

Sorry this post wasn't as amusing as its title would suggest; I shall try harder next time.

No comments:

Post a Comment