Helping People Sketch Fast with Computer-Aided Chunking

Here's an idea I'd like to try to help myself and other artists who want to sketch more quickly, for example for sketching people passing by during "urban sketching".

1) Take lots of photos of people in an environment similar to the one you'll be drawing in.

2) Edge-detect them. The results would look rather like this.

3) Edit away parts of the images that aren't relevant to the step below.

4) Feed the results into a program that searches for line shapes that are common to a lot of the images. For example, the rough "K"-shape made by the backs of the legs of someone standing at ease at a counter, one leg straight and one bent. The straight leg makes the straight part of the K, and the bent leg the "<" part.

5) Process the resulting list of line shapes so that they grab the attention as much as possible. Perhaps by increasing their contrast, orcolouring them yellow or red.

6) Train the artist on the results.

The idea is to equip the artist's brain with a repertoire of shapes that often occur in the scenes being drawn. This could be useful in at least two ways. First, it might help them remember the scene. When trying to memorise what I'm seeing so that I can draw it later, I find that I consciously search for such shapes and then try to vary or "modulate" them so that they match the appropriate part of the scene exactly.

Second, it might help them draw the scene more quickly, if the artist has practised drawing these shapes as well as memorising them.

I suspect that most artists do build up a reportoire of such shapes without conscious effort, as they continue to draw. But acquiring them in that way is haphazard: surely we can make doing so more efficient.

What I'm basing my idea on is the well-known psychological phenomenon of chunking. After writing the first version of this post, I discovered that there has indeed been research into drawing and chunking. One of the papers I found was "Graphical Production of Complex Abstract Diagrams: Drawing Out Chunks and Schemas" by Unaizah Obaidellah and Peter Cheng. They asked subjects to draw a variant of the Rey-Osterrieth Complex Figure: by tracing, by copying, by drawing immediately from memory, and by drawing from memory after a delay. All their subjects appeared to organise the figure being drawn into chunks for all the drawing tasks, even tracing.

Trans-Cranial Magnetic Stimulation and the Curled-Cat and Bristly-Scottie Drawings of Allan Snyder's Subjects

I thought I'd say some more about trans-cranial magnetic stimulation (TMS) and drawing, which I mentioned here. The paper that got me interested is "Savant-like Skills Exposed in Normal People by Suppressing the Left Fronto-Temporal Lobe", Journal of Integrative Neuroscience, 2, 2, 2003, by Allan Snyder and colleagues. The cat and dog drawings shown at the top of page 4 excited me, and I'll explain why next.

Many art teachers say that novices draw objects "symbolically", usually showing the object's main parts, from a point of view that makes them instantly recognisable. Novices often draw objects "flat" as well, side-on or face-on. Well-known examples are spectacles drawn as a pair of circles, and trees drawn as a trunk with branches sticking out flat on either side.

That's true of the non-TMS cats drawn by Snyder's subject N.R. and the dogs drawn by A.J. But what excites me is that under TMS, N.R.'s cat changed from side-on to curled round on itself. That's a complicated posture that many novices probably wouldn't be able to recall, and wouldn't dare to draw even if they could. Was the TMS somehow causing N.R. to access a previously inacessible memory?

Likewise with A.J.'s dogs. Because of foreshortening, a dog's muzzle is difficult to draw face-on. But something has prompted A.J. to do so under TMS, and perhaps even to draw the bristles around the muzzle receding back into the distance.

What has this to do with TMS? Snyder hypothesises that savants' remarkable skills are innate, rather than arising through intensive practice or greater development of parts of the brain. These skills, he says, work by accessing low-level information that's in everyone's brains — in the case of drawing savants such as Stephen Wiltshire, an analogue representation of the visual field. Normal people can't access such information, but savants can because of their brain impairments, so if we can simulate these impairments via TMS, perhaps we can induce savant skills in normal people.

Snyder doesn't go into detail about the neural mechanisms, and some neuroscientists I've spoken to have criticised him for this. But I think this research needs to be regarded in the same way as the discovery of drugs such as opium and digitalis. The mechanisms may not be known, but it's obvious that something interesting and useful is happening, and that it's worth working out how to enhance it.